Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Glare studies: Comparison of three glare indices, HDR imaging and measured values
(USC Thesis Other)
Glare studies: Comparison of three glare indices, HDR imaging and measured values
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
GLARE STUDIES: COMPARISON OF THREE GLARE INDICES, HDR IMAGING AND MEASURED VALUES by Hanshu Yin A Thesis Presented to the FACULTY OF THE USC SCHOOL OF ARCHITECTURE UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree MASTER OF BUILDING SCIENCE May 2011 Copyright 2011 Hanshu Yin ii Acknowledgments: I would like to express my acknowledgements to my committee members Prof. Schiler, Prof. Woll and Suk. They lead me and gave me tremendous help, support and inspiration during the whole process. And I need to express my sincere gratefulness to Greg Ward, who helped me solved countless problems in the thesis with his knowledge, kindness and patience. Last but not least, I would like express my special gratitude to my parents, professors, colleagues and friends for all the support they gave me. iii Table of Contents: Acknowledgments: ............................................................................................................. ii List of Tables: .................................................................................................................... vi List of Figures: ................................................................................................................... ix Abstract: ........................................................................................................................... xvi Chapter 1: Introduction of Research Background .............................................................. 1 1.1 Light .......................................................................................................................... 2 1.1.1 Physics of Light ................................................................................................. 2 1.2 Vision and Visual Performance ................................................................................ 9 1.2.1 Light and Vision ................................................................................................ 9 1.2.2 Light and Visual Performance ......................................................................... 12 1.3 Light Sources and Lamp Types .............................................................................. 14 1.3.1 Light Sources ................................................................................................... 15 1.3.2 Lamp Types ..................................................................................................... 18 1.4 Glare ........................................................................................................................ 22 1.4.1 Definitions of Glare ......................................................................................... 22 1.4.2 Glare Categories ............................................................................................... 22 1.5 Glare Analysis Methods .......................................................................................... 24 1.5.1 Psychological Statistical Survey ...................................................................... 24 1.5.2 Visual Comfort Probability .............................................................................. 25 1.5.3 The Luminance Histogram Method ................................................................. 26 1.5.4 Calculated Glare Indices .................................................................................. 27 Chapter 2: Glare Analysis ................................................................................................. 31 2.1 High Dynamic Range Image ................................................................................... 32 2.1.1 Digital Camera and Camera Response Curve .................................................. 32 2.1.2 High Dynamic Range Image Software & Format ............................................ 37 2.2 HDR analysis software............................................................................................ 40 2.2.1 Radiance ........................................................................................................... 40 2.2.2 GlareIndices (C++ Program) ............................................................................ 40 Chapter 3: HDR Experiments ........................................................................................... 46 3.1 Experiment Test 1: Camera (with Photosphere) ..................................................... 47 3.2 Experiment Test 2: Reflectance .............................................................................. 77 3.3 Experiment Test 3: Software .................................................................................. 87 iv Chapter 4: Three Glare Analysis by Using HDR Images ............................................... 107 4.1 Experiment Model Tests ....................................................................................... 107 4.1.2 GlareIndices (C++ Program) Glare Analysis ................................................. 114 4.1.3 Radiance Glare Analysis ................................................................................ 117 4.1.4 Comparison between GlareIndices (C++ Program) and Radiance Results ... 119 4.1.5 Comparison among Three Glare Indices: BGI, UGR and Schiler Index ....... 120 4.2 Interior Test ........................................................................................................... 121 4.2.1 Interior Settings .............................................................................................. 121 4.2.2 GlareIndices(C++ Program) Glare Analysis .................................................. 124 4.2.3 Radiance Glare Analysis ................................................................................ 125 4.2.4 Comparison between GlareIndices (C++ Program) and Radiance Results ... 126 4.2.5 Comparison among Three Glare Indices: BGI, UGR and Schiler Index ....... 126 4.3 Exterior Night Test 1 ............................................................................................ 127 4.3.1 Test Settings ................................................................................................... 127 4.3.2 GlareIndices (C++ Program) Glare Analysis ................................................. 129 4.3.3 Radiance Glare Analysis ................................................................................ 130 4.3.4 Comparison GlareIndices (C++ Program) and Radiance Results .................. 131 4.3.5 Comparison among Three Glare Indices: BGI, UGR and Schiler Index ....... 131 4.4 Exterior Test 2 ....................................................................................................... 132 4.4.1 Test Settings ................................................................................................... 132 4.4.2 GlareIndices (C++ Program) Glare Analysis ................................................. 134 4.4.3 Radiance Glare Analysis ................................................................................ 136 4.4.4 Comparison GlareIndices (C++ Program) and Radiance Results .................. 136 4.4.5 Comparison among Three Glare Indices: BGI, UGR and Schiler Index ....... 137 4.5 Analysis on Glare Analysis by Using HDR Images ............................................. 137 4.5.1 Comparison between GlareIndices (C++ Program) and Radiance ................ 137 4.5.2 Analysis of HDR Images as a Glare Analysis Tool ....................................... 138 4.5.3 Glare Indices UGR and Luminance Histogram Method ................................ 140 Chapter 5: Conclusion..................................................................................................... 142 5.1 HDR Image as a Luminance Mapping Tool ......................................................... 142 5.2 Glare Analysis Based on HDR images ................................................................. 145 5.2.1 GlareIndices (C++ Program) .......................................................................... 145 5.2.2 GlareIndice (C++ Program) BGI and UGR ................................................... 146 5.2.3 GlareIndices (C++ Program) Luminance Histogram Method ....................... 146 Chapter 6: Future Work .................................................................................................. 148 v References ....................................................................................................................... 150 Appendix: Codes ............................................................................................................. 152 vi List of Tables: Table 1: Typical Glare Criterion and BRS ....................................................................... 30 Table 2: Typical Glare Criterion and UGR ....................................................................... 30 Table 3: Test 1 Cameras Settings...................................................................................... 48 Table 4: Luminance Meter Readings of Each Measured Block ....................................... 49 Table 5: Canon Rebel T2i HDR Image Luminance Readings from Photosphere ............ 56 Table 6: Luminance Ratios (over White Surface) from Luminance Meter ...................... 57 Table 7: Luminance Ratios (over White Surface) from Canon T2i HDR ........................ 58 Table 8: Luminance Ratios (over White Surface) from Panasonic ZS1 HDR ................. 59 Table 9: Comparison of White Surface Canon T2i HDR and Luminance Meter ............. 61 Table 10: Comparison of Light Gray Surface Canon T2i HDR and Luminance Meter ... 62 Table 11: Comparison of Dark Gray Surface Canon T2i HDR and Luminance Meter ... 63 Table 12: Comparison of Black Surface Canon T2i HDR and Luminance Meter ........... 64 Table 13: The Errors of Measured Surfaces under Different Light Levels ...................... 65 Table 14: Panasonic HDR Image Luminance Readings from Photosphere ..................... 68 Table 15: Comparison of White Surface Panasonic ZS1 HDR and Luminance Meter .... 70 Table 16: Comparison of Light Gray Surface Panasonic HDR and Luminance Meter .... 71 Table 17: Comparison of Dark Gray surface Panasonic HDR and Luminance Meter ..... 72 vii Table 18: Comparison of Black Surface Panasonic HDR and Luminance Meter ............ 73 Table 19: Panasonic Luminance Reading Errors under Different Light Levels ............... 74 Table 20: Gretage MacBeth Basic Information ................................................................ 79 Table 21: Test 2 Camera Settings ..................................................................................... 79 Table 22: Test 2 Luminance Meter Readings ................................................................... 81 Table 23: Canon T2i HDR Image Luminance Readings .................................................. 82 Table 24: Test 2 Canon T2i HDR luminance reading errors ............................................ 84 Table 25: Luminance Meter Readings .............................................................................. 89 Table 26: Canon G12 HDR Luminosity from Adobe Photoshop ..................................... 90 Table 27: Calibrated Luminance Reading of Canon G12 HDR Images ........................... 90 Table 28: Canon G12 Photoshop Luminance Reading Error (Calibration at Block 1) .... 91 Table 29: Calibrated Luminance Reading of Canon G12 HDR Images ........................... 93 Table 30: Canon G12 Photoshop Luminance Reading Error (at Block 6) ....................... 94 Table 31: Canon T2i Bracket HDR Luminance Readings ................................................ 96 Table 32: Calibrated Canon T2i Bracket HDR Luminance Readings (at Block 1) .......... 97 Table 33: Calibrated Canon T2i Bracket HDR Luminance Readings Errors ................... 97 Table 34: Calibrated Canon T2i Bracket HDR Luminance Readings ............................ 100 Table 35: Calibrated Canon T2i Bracket HDR Luminance Readings Errors ................. 100 Table 36: Canon T2i Photosphere HDR Luminance Readings ...................................... 102 viii Table 37: Calibrated Canon T2i Photosphere HDR Luminance Readings ..................... 103 Table 38: Calibrated Canon T2i Photosphere HDR Luminance Readings Errors .......... 103 Table 39: Non-calibrated Canon T2i Photosphere HDR Luminance Readings Errors .. 104 Table 40: Canon T2i Settings ......................................................................................... 109 Table 41: Canon T2i Settings ......................................................................................... 122 Table 42: Canon T2i settings .......................................................................................... 127 ix List of Figures: Figure 1: Electromagnetic Spectrum of Light .................................................................... 3 Figure 2: Young's Diffraction of Light ............................................................................... 4 Figure 3: Specular Reflection ............................................................................................. 5 Figure 4: Diffused Reflection ............................................................................................. 5 Figure 5: Diffused Transmission ........................................................................................ 6 Figure 6: Non-diffuse Transmission ................................................................................... 6 Figure 7: Photo of Konica Minolta Luminance Meter 1° ................................................... 7 Figure 8: Photo of IALD 61-680 Light Meter .................................................................... 8 Figure 9: Illustration of Vision ........................................................................................... 9 Figure 10: Field of Vision of a Normal Pair of Human Eyes ........................................... 11 Figure 11: Subtended Angle of Eyes ................................................................................ 11 Figure 12: Vision Field ..................................................................................................... 11 Figure 13: Human Eyes Perceivable Luminance Range ................................................... 13 Figure 14:Deco 400W Med 120V G30 CL 1CT .............................................................. 18 Figure 15:20W GU5.3 12V MR16 10D 3BC ................................................................... 19 Figure 16:PL-C ALTO 13W/827 1CT .............................................................................. 19 Figure 17:SOX 135W 1SL................................................................................................ 20 x Figure 18: MASTER SON-T PIA Plus 400W Mogul 220V 1SL ..................................... 20 Figure 19: MHN-SA 1800W/956 230V TD40 ................................................................. 21 Figure 20: Direct Glare and Reflected Glare .................................................................... 24 Figure 21: CCD Sensor ..................................................................................................... 33 Figure 22: Bayer Arrangement of Color Filters ................................................................ 35 Figure 23: Foveon X3 Sensor ........................................................................................... 35 Figure 24: Nikon's Dichroic Color Seperation Patent Drawing ....................................... 35 Figure 25: GlareIndices (C++ Program) Interface ............................................................ 43 Figure 26: Warning Dialog in Step 4 ................................................................................ 44 Figure 27: Input BellMax, BellMin, SpikeMax and SpikeMin in Step 7 ......................... 45 Figure 28: A Demonstration of Final Result Output ........................................................ 45 Figure 29: Illuminance Meter Reading ............................................................................. 48 Figure 30: Panasonic -2EV Exposure ............................................................................... 51 Figure 31: Panasonic -1EV Exposure ............................................................................... 51 Figure 32: Panasonic 0EV Exposure ................................................................................ 51 Figure 33: Panasonic +1EV Exposure .............................................................................. 51 Figure 34: Panasonic +2EV Exposure .............................................................................. 51 Figure 35: Photosphere HDR Image ................................................................................. 52 Figure 36: Panasonic -2EV Exposure ............................................................................... 54 xi Figure 37: Panasonic -1EV Exposure ............................................................................... 54 Figure 38: Panasonic 0EV Exposure ................................................................................ 54 Figure 39: Panasonic -+1EV Exposure ............................................................................. 54 Figure 40: Panasonic +2EV Exposure .............................................................................. 54 Figure 41: Test 1 Panasonic Photosphere HDR Image ..................................................... 55 Figure 42: Overall Comparison between Canon T2i HDR and Luminance Meter .......... 60 Figure 43: Comparison of White Surface Canon T2i HDR and Luminance Meter ......... 61 Figure 44: Comparison of Light Gray Surface Canon T2i HDR and Luminance Meter . 62 Figure 45: Comparison of Dark Gray Surface Canon T2i HDR and Luminance Meter .. 63 Figure 46: Comparison of Black Surface Canon T2i HDR and Luminance Meter .......... 64 Figure 47: Canon T2i HDR Luminance Error ( X axis is logarithm, log2) ...................... 67 Figure 48: Canon T2i HDR Luminance Errors ................................................................. 67 Figure 49: Overall Comparison between Panasonic HDR and Luminance Meter ........... 69 Figure 50: Comparison of White Surface Panasonic HDR and Luminance Meter .......... 70 Figure 51: Comparison of Light Gray Surface Panasonic HDR and Luminance Meter .. 71 Figure 52: Comparison of Dark Gray Surface Panasonic HDR and Luminance Meter ... 72 Figure 53: Comparison of Black Surface Panasonic HDR and Luminance Meter ........... 73 Figure 54: Panasonic HDR Luminance Error ( X axis is logarithm, log 2)...................... 75 Figure 55: Panasonic HDR Luminance Error ................................................................... 76 xii Figure 56: Gretag MacBeth Color Checker ...................................................................... 78 Figure 57: -2EV Exposure Photo of Illuminance at 4.9fc ................................................ 80 Figure 58: 0EV Exposure Photo of Illuminance at 4.9fc .................................................. 80 Figure 59:+2EV Exposure Photo of Illuminance at 4.9fc ................................................ 81 Figure 60: Overall Comparison between Canon T2i HDR and Luminance Meter .......... 83 Figure 61: Canon T2i HDR Luminance Error ( X axis is logarithm, log 2) ..................... 86 Figure 62: Canon T2i HDR Luminance Reading Error .................................................... 87 Figure 63: Canon G12 HDR Image Illuminance at 0.7 fc ................................................ 88 Figure 64: Overall Comparison between Canon G12 HDR and Luminance Meter ......... 92 Figure 65: Canon G12 HDR Luminance Reading Error (Calibration at Block 1) ........... 92 Figure 66: Overall Comparison between Canon G12 HDR and Luminance Meter ......... 95 Figure 67: Canon G12 HDR Luminance Reading Error (Calibration at Block 6) ........... 95 Figure 68: Overall Comparison of Calibrated Bracket HDR and Luminance Meter ....... 98 Figure 69: Canon T2i Bracket HDR Luminance Reading Error ...................................... 99 Figure 70: Overall Comparison of Calibrated Bracket HDR and Luminance Meter ..... 101 Figure 71: Canon T2i Bracket HDR Luminance Reading Error .................................... 102 Figure 72: Comparison of Three HDR Image Software ................................................. 106 Figure 73: The Front Side (with Windows) of the Model .............................................. 108 Figure 74: The Back Side of the Model and Camera is Mounted on a Tripod ............... 108 xiii Figure 75: Equivalent to -6EV ........................................................................................ 110 Figure 76: Equivalent to -5EV ........................................................................................ 110 Figure 77: Equivalent to -4EV ........................................................................................ 110 Figure 78: Equivalent to -3EV ........................................................................................ 110 Figure 79: Equivalent to -2EV ........................................................................................ 110 Figure 80: Equivalent to -1EV ........................................................................................ 110 Figure 81: Equivalent to 0EV ......................................................................................... 111 Figure 82: Equivalent to +1EV ....................................................................................... 111 Figure 83: Equivalent to +2EV ....................................................................................... 111 Figure 84: Equivalent to -10EV ...................................................................................... 113 Figure 85: Equivalent to -8EV ........................................................................................ 113 Figure 86: Equivalent to -6EV ........................................................................................ 113 Figure 87: Equivalent to -4EV ........................................................................................ 113 Figure 88: Equivalent to -2EV ........................................................................................ 113 Figure 89: Equivalent to 0EV ......................................................................................... 113 Figure 90: Equivalent to +2EV ....................................................................................... 114 Figure 91: Test 1 Histogram in Excel ............................................................................. 115 Figure 92: Test 2 Hitogram in Excel ............................................................................... 116 Figure 93: Test 1 Radiance Screenprint .......................................................................... 118 xiv Figure 94: Test 2 Radiance Screenprint .......................................................................... 119 Figure 95: Equivalent to -7EV ........................................................................................ 123 Figure 96: Equivalent to -5EV ........................................................................................ 123 Figure 97: Equivalent to -3EV ........................................................................................ 123 Figure 98: Equivalent to -1EV ........................................................................................ 123 Figure 99: Equivalent to +1EV ....................................................................................... 123 Figure 100: Equivalent to +3EV ..................................................................................... 123 Figure 101: Equivalent to +5EV ..................................................................................... 124 Figure 102: Equivalent to +7EV ..................................................................................... 124 Figure 103: Interior Test Histogram in Excel (Y is scaled at log2) ................................ 124 Figure 104: Interior Test Radiance Screenprint .............................................................. 125 Figure 105: -2EV ............................................................................................................ 128 Figure 106: -1EV ............................................................................................................ 128 Figure 107: 0EV .............................................................................................................. 128 Figure 108: +1EV ........................................................................................................... 128 Figure 109: +2EV ........................................................................................................... 128 Figure 110: Exterior Night Test 1 Histogram in Excel (Y axis is scaled at log2) .......... 129 Figure 111: Exterior Night Test 1 Radiance Screenprint ................................................ 130 Figure 112: Equivalent to -11EV .................................................................................... 133 xv Figure 113: Equivalent to -9EV ...................................................................................... 133 Figure 114: Equivalent to -7EV ...................................................................................... 133 Figure 115: Equivalent to -5EV ...................................................................................... 133 Figure 116: Equivalent to -3EV ...................................................................................... 133 Figure 117: Equivalent to -1EV ...................................................................................... 133 Figure 118: Equivalent to +1EV ..................................................................................... 134 Figure 119: Equivalent to +3EV ..................................................................................... 134 Figure 120: Exterior Night Test 2 Histogram in Excel (both in logarithm) ................... 134 Figure 121:Exterior Night Test 2 Histogram in Excel (Linear) ...................................... 135 Figure 122: Exterior Night Test 2 Radiance Screenprint ................................................ 136 xvi Abstract: People’s perception of glare is primarily based on two factors: absolute brightness and contrast. One possible glare analysis method is using empirical formulas to predict people’s feeling of the possible magnitude of un-comfortableness. High Dynamic Range image (HDR) is a type of processed digital image that contains much higher dynamic range information which could potentially be used a luminance mapping tool. In the thesis, author tested several factors which could affect the accuracy of luminance mapping by HDR image and what the extent of influence they are. By using HDR images as luminance maps, the author proposed a C++ program GlareIndices to calculate three glare indices namely Unified Glare Rating, BRS glare equation and Luminance Histogram Ratio to predict glare possibility in an existing scene without involving human subjects. . 1 Chapter 1: Introduction of Research Background Light is the fundamental necessity for human beings’ visual experience. In architecture, light is useful to reveal the forms and shapes of a design, both interior and exterior. Architectural lighting designers as well as architects generally design artificial and natural lighting systems to achieve aesthetic values. On the other hand, electrical engineers and lighting designers consider lighting systems as a tool to help people within the space to obtain an adequate and functional visual environment for tasks such as studying, drawing, surgical operation, etc. In either case the quality of lighting design is dependent on the designers’ concerns on light levels, use of colors, fixture locations and positions, fixture types, lamp types, psychological comfort of the people within the environment, energy consumption, controllability, economic efficiency and maintenance. Among all these factors, psychological comfort of the people within the environment is one of the most importance elements, which is largely interrelated with the other factors. It has significant influence on people’s experience of space, for example, the openness of the space, thermal comfort of the space and safety and hygiene of the space all of which may cause consequential impacts on people’s mood and productivity. Among all the 2 lighting design related complains, insufficient light levels and glare are the most crucial problems. People’s perception of glare is primarily based on two factors: absolute brightness and contrast. In this thesis, the author is trying to analysis glare by using empirical formulas to predict people’s feeling of the possible magnitude of un-comfortableness. 1.1 Light 1.1.1 Physics of Light From a physical viewpoint, light can be regarded as that portion of the electromagnetic spectrum which lays between the wavelengths limits of 380 nanometers and 780 nanometer (John E. Kaufman 1978, p2-1). It is illustrated in Figure 1 (Wiki n.d.) 3 Figure 1: Electromagnetic Spectrum of Light Within the light spectrum range we define colors based on the wavelength. From the shortest wavelength, highest frequency, the color spectrum is violet (wavelength: 380-450 nm, frequency: 668-789 THz), blue (wavelength: 450-475 nm, frequency: 631-668 THz), cyan (wavelength: 476-495 nm, frequency: 606-630 THz), green (wavelength: 495-570 nm, frequency: 508-526 THz), yellow (wavelength: 570-590 nm, frequency: 508-526 THz), orange (wavelength: 590-620 nm, frequency: 484-508 THz), red (wavelength: 620-750 nm, frequency: 400-484 THz) (Wiki, Visible Spectrum n.d.) Light has both wave and particle properties. This theory was devised by Albert Einistein in the early 1900s. The wave property determines that light waves can interfere with each other like water waves or sound waves, and it can also be polarized. The particle 4 properties determine that light can be absorbed and emitted only in discrete package called photon. It is illustrated in Figure 2 (Wiki, Young's Diffraction of Light n.d.) Figure 2: Young's Diffraction of Light 1.1.2 Terminology and Definition Reflectance/reflectance factor/reflectance coefficient: the ratio of reflected light to incident light. If the reflection happens on a polished surface such as a gloss aluminum sheet or glass, it is called specular reflection, as in Figure 3: Specular Reflection (Stein 2006, P462) If the surface is coarse, reflections occurs on the many small surface projections and the light is diffused, as in Figure 4: Diffused Reflection. (Stein 2006, P462) 5 Figure 3: Specular Reflection Figure 4: Diffused Reflection Reflectance can be measured by the Known Sample Comparison Method or the Reflected/Incident Light Method. (Stein 2006, P461)Both of the two methods are measuring combined specular and diffused reflection ratio on the surface. Luminous Transmittance/transmittance/transmission factor/coefficient of transmission: is a measure of a material’s capability to transmit incident light, which is the ratio of the total transmitted light to the total incident light. (Stein 2006, P465) Some materials may exhibits spectrum selective transmission or, in other words, spectrum selective absorption. Similar to the definition of reflectance, luminous transmittance also has diffusing and non-diffusing aspects, as in Figure 5: Diffuse Transmission (Stein 2006, P461) and Figure 6: Diffuse Transmission (Stein 2006, P461). 6 Figure 5: Diffused Transmission Figure 6: Non-diffuse Transmission Luminous Flux: is the measure of power (or rate of energy) emitted from a source in the visible wavelength range, which is 380nm to 780nm. The unit of luminous flux is the lumen (lm) which is defined in terms of candela (cd). (J. R. Coaton 1997, P18) 1 lumen is defined as 1 cd light source radiates light evenly in all directions within a transparent sphere of 1 ft radius, and the luminous energy emanating from 1 square ft of the surface of sphere is 1 lumen. (Stein 2006, P465)One watt of radiant power at 555 nm - the wavelength at which the typical human eye is most sensitive - is equivalent to a luminous flux of 680 lumens. (Halsted 1993) Luminous Intensity: is the luminous flux per solid angle emitted or reflected from a point. (Halsted 1993) The unit of luminous intensity is the lumen per steradian (lm/ sr), or 7 candela (cd). Luminous intensity cannot be measured directly by any single device, but it can be calculated from its illuminance effects. Luminance: is the luminous intensity per unit area of apparent (projected) area of primary (emitting) or secondary (reflecting) light source in a given direction. (Stein 2006, P465)The unit of luminance is candela per square meter (cd/ m 2 ) in SI units, sometimes also named nit and is similar to footlambert (fL), in I-P units. The footlambert is equal to 1 / π candela per square foot. 1 fL= 3.426 cd/m 2 Luminance can be measured using a luminance meter, as in Figure 7: Photo of Konica Minolta Luminance Meter 1° Figure 7: Photo of Konica Minolta Luminance Meter 1° Illuminance: is the density of luminous power, it is the luminous flux incident on a surface per unit area. (Halsted 1993) The unit of illuminance is lux in SI unit which is defined as 1 lumen of luminous flux uniformly incident perpendicular to the surface of 8 1 m 2 area; therefore, 1fc= 10.764 lux. Illuminance can be measured by a light meter/ illuminance meter, which is usually made up of a photoelectric material, microammeter and electronic control circuitry. Most light meters are color and cosine corrected. Color correction is required because the human eye’s differential sensitivity to colors/ light wavelength is different than the inherent spectral sensitivity of the light meter. By color correction, the light meter’s spectrum response is made to correspond with that of human eyes’. Cosine correction is provided so that such effects as reflection from the glass or shielding from the meter housing do not cause misreading of light coming from off-axis.. Figure 8: Photo of IALD 61-680 Light Meter Figure 8: Photo of IALD 61-680 Light Meter 9 1.2 Vision and Visual Performance 1.2.1 Light and Vision Light enters the eye through the pupil. The amount of light getting through pupil is controlled by the size of pupil which is controlled by the iris. The lens behind the pupil then focuses the image onto the retina. The optic nerve conveys the received information to the brain by electrical impulses. A vision is created in the brain thorough this process. (Stein 2006, P474), Figure 9: Illustration of Vision (Illustration of Vision n.d.). Figure 9: Illustration of Vision Rod and cone cells are the types of light-sensitive cells on the retina. Cone cells give us the sensation of color and details. They are capable of response to luminance ranging from 3 to1,000,000 cd/m 2 . Rod cells are extremely sensitive to light but do not have color sensitivity. They are capable of response to the luminance in the range 1/1000 to120 10 cd/m 2 . Rod cells are slower acting than cone cells and lacking detail discrimination but they are extraordinary motion sensitive. (Stein 2006, P474) Vision Field: Human eyes have very broad vision field, but the field is not unified. Different functions happen at different areas of the vision field. The acute perception of detail vision is called the extremely narrow cone of foveal (central) vision. The coarser sight information is collected by the near field or surround which is called binocular vision, is about 30° half-angle, located around the foveal cone. The farthest vision area is called the field and peripheral area, which primarily gives us our subjective, ambience-type reactions (Stein 2006, P474). Because of this organization of human eyes, we have a 180-degree wide forward-facing field of view and vertically 130-degree and horizontally around 130-degree binocular vision, as illustrated in Figure 10 (Stein 2006, P475) , Figure 11 (Stein 2006, P475) and Figure 12 (Vision Field n.d.). 11 Figure 10: Field of Vision of a Normal Pair of Human Eyes Figure 11: Subtended Angle of Eyes Figure 12: Vision Field 12 1.2.2 Light and Visual Performance Brightness and Adaptation Human eyes are capable of perceiving luminance range of more than 100,000,000 to 1, as in Figure 13 (Halsted 1993), which is a very large range. However, for low light levels, human eyes need a period of time, called adaptation time, to accommodate. As mentioned earlier, the two types—rod and cone, of sensor cells in the eyes have different response light magnitudes for this reason, several modes of perception are distinguished. They are respectively: Photopic Vision for light-adapted eyes, Scotopic Vision for dark-adapted eyes and Mesopic Vision just in between. 13 Figure 13: Human Eyes Perceivable Luminance Range Contrast, Adaptation and Visual Acuity Visual acuity is largely dependent on the contrast of the brightness across the task area. However, the contrast of brightness between the observed object and its background may also affect the visual acuity. When dealing with full-spectrum white light and ignoring chromaticity, contrast is the single most important factor in visual acuity (Stein 2006, 14 P476). Contrast is a dimensionless ratio of the differential between luminance of the task object plus background luminance divided by background luminance. For a maximum visual acuity, the luminance of a surface-type task should be the same as, or slightly higher than that of the background, but a ratio of 3:1 can be accepted for most situations (Stein 2006, P476). Visual acuity increases with increased adaptation level. Color and color contrast of objects may influence visual acuity in a very complex way. As previous studies demonstrate, the subjective brightness of a colored object is greater than that of an achromatic object for the same photometric luminance. Contrast adaptation and visual performance under a not-full spectrum white light source, are still undergoing research. 1.3 Light Sources and Lamp Types Light source selection and design is among the most important design elements for luminous environment. The emission of visible light by a material after it has been exposed to ultraviolet light (UV) or some other method of excitation is defined as luminescence (John E. Kaufman 1978). Fluorescence is the term for emission which happens immediately after excitation, and phosphorescence is for the circumstance of a delay or long afterglow. 15 1.3.1 Light Sources Natural Phenomena Sunlight: is the energy of color temperature about 6500K received from the sun. The average luminance of the sun is approximately 160,000 candelas per square centimeter (1.6× 10 9 cd/m 2 ) viewed from sea level. (John E. Kaufman 1978, 2-6) Sky light: is the light which scattered in all directions by the earth’s atmosphere. The blue clear sky, white clouds and reddish appearance of the sky in the morning are all examples of scattering effects. The sky luminance is around 3,500 candelas per square centimeter (3.5× 10 7 cd/m 2 ) (John E. Kaufman 1978, 2-6). Moonlight: is the light from sun reflected to earth by the moon. The luminance of the moon is around 0.25 candelas per square centimeter (John E. Kaufman 1978, 2-6). Lightning: is a sudden release in a spark type of discharge of accumulated electrical charges in clouds. The lighting spectrum consists principally of nitrogen bands (John E. Kaufman 1978, 2-6). Aurora Borealis and Aurora Australis: are caused by electron streams spiraling into the atmosphere (John E. Kaufman 1978, 2-6). 16 Bioluminescence: is a form of chemi-luminescence. When certain compounds which are manufactured by plants and animals are oxidized, they produce light. Bioluminescent sources can have very high lumen per watt ratings (John E. Kaufman 1978, 2-6) Artificial Light Incandescence: 1. Filament: the heating effect of electric current raises the temperature of the filament causing the electronic transitions in atoms and molecules which occur with emmitance of light (John E. Kaufman 1978, 2-6). 2. Pyroluminescence: is also known as flame luminescence, light generated by the combustion process which is energy exchange between highly excited molecules and atoms under high temperature (John E. Kaufman 1978, 2-6). 3. Candoluminescence: is also known as gas mantle, the fluorescence excited by incandescent radiation of certain types of materials (John E. Kaufman 1978, 2-6). 4. Arc Crater: is light produced by incandescence of the electrodes (crater) and the luminescence of vaporized electrode material and the surrounding gaseous atmosphere (John E. Kaufman 1978, 2-6). 17 Luminescence: 1. Gaseous Discharge: is the light produced when electrical discharge stimulates light emission in a gas (John E. Kaufman 1978, 2-6). 2. Photoluminescennce: is visible light emitted after absorption of a photon of visible ultraviolet radiation (John E. Kaufman 1978, 2-6). 3. Electoluminescence: is light directly converted from current energy without using any intermediate step of ultraviolet radiation (John E. Kaufman 1978, 2-6). 4. Galvanoluminescence: is light which appears at either the anode or cathode when solutions are electrolyze (John E. Kaufman 1978, 2-6). 5. Crystalloluminescence: is light produced when solutions crystallize (John E. Kaufman 1978, 2-6). 6. Chemiluminescence: is light produced during a chemical reaction at room temperature (John E. Kaufman 1978, 2-6). 7. Triboluminescence: is light produced by shaking, rubbing or crushing crystals (John E. Kaufman 1978, 2-6). 8. Sonoluminescence: is light observed when sound waves are passed through fluids (John E. Kaufman 1978, 2-6). 9. Radioluminescence: is light emitted from a material which is under bombardment from alpha rays, beta rays, gamma rays or X rays (John E. Kaufman 1978, 2-6). 18 1.3.2 Lamp Types Incandescent Lamp: Electric current passing through a coiled filament wire generates heat, raising the temperature to the point where blackbody radiation occurs in the visible as well as in the infrared. It has low luminous efficacy, but has a full-spectrum color quality with larger amount of light in long wavelength producing a yellowish- red tone (John E. Kaufman 1978, 8-1). Figure 14 (Philips n.d.). Figure 14:Deco 400W Med 120V G30 CL 1CT Tungsten Halogen Lamp: Electric current passing through the filament generates heat raising the temperature to the point where blackbody radiation occurs in the visible as well as in the infrared. It has a small capsule made from quartz glass and filled with mixture of halogen gas and inert gas around the filament which allows the filament temperature to be raised to a higher level than the typical incandescent lamp. This produces a higher luminous efficacy and a whiter light with similar characteristics of continuous, spectral distribution (John E. Kaufman 1978, 8-1). Figure 15 (Philips n.d.). 19 Figure 15:20W GU5.3 12V MR16 10D 3BC Fluorescent Lamp: electric-discharge through low-pressure mercury-vapor converts electrical energy into electromagnetic radiation which mainly lies in ultraviolet (UV) region. The UV radiation is absorbed and converted to visible light when it hit the phosphor coat on the inner side of lamp. The source has good luminous efficacy especially when certain high-efficiency phosphor can be selected to convert the UV radiation to 555 nm light output which is in the maximum visual sensitive spectrum to achieve extremely high efficacy. Reasonably white light can be generated by applying correct proportions of phosphors (John E. Kaufman 1978, 8-17). Figure 6 (Philips n.d.). Figure 16:PL-C ALTO 13W/827 1CT 20 Low-pressure Sodium Lamp: this source uses electrical discharge through low pressure sodium gas. The source is virtually monochromatic, and has the highest efficacy (John E. Kaufman 1978, 8-43). Figure 17 (Philips n.d.). Figure 17:SOX 135W 1SL High-pressure Sodium Lamp: its arc tube body is made from a special ceramic material and Xenon is the start gas and then sodium and mercury form an amalgam acting as the stable lamp vapor. Because of the light generation method, a high-pressure sodium lamp has high luminous efficacy with limited distribution of radiation wavelength range (John E. Kaufman 1978, 8-36). Figure 18 (Philips n.d.). Figure 18: MASTER SON-T PIA Plus 400W Mogul 220V 1SL 21 Metal Halide Lamp: is another gas-discharge (high-intensity discharge - HID) source. It has full spectrum and good color rendering as well as luminous efficacy but has a poor color consistency over its lifetime. It requires a high starting voltage (John E. Kaufman 1978, 8-37). Figure 19 (Philips n.d.). Figure 19: MHN-SA 1800W/956 230V TD40 Light-Emitting Diode (LED) Lamp: this source produces light by passing electric current through a semiconductor diode. The color of emitted light is dependent on the band gap material of the semiconductor. LEDs have high luminous efficacy and are available in the entire color range (J. R. Coaton 1997, P281). 22 1.4 Glare 1.4.1 Definitions of Glare 1. The sensation produced by luminance within the visual field that is sufficiently greater than the luminance to which the eyes are adapted to cause annoyance, discomfort, or loss in visual performance and visibility. (John E. Kaufman 1978) 2. Excessive luminance and /or excessive luminance ratios in the field of vision are commonly referred to as glare (Stein 2006, P490) In both definitions, glare is a visual sensation. It is caused by absolute extreme luminance or excessive contrast between certain light sources and overall luminance in the vision field. The magnitude of the sensation of glare varies with several factors such as the light sources positions, sizes, quantity and the adapted luminance of the eyes. 1.4.2 Glare Categories By luminance level: 1. Absolute Glare: There is an extremely high luminance source presented in the view field, and this light source is causing discomfort even if there is no severe brightness contrast within the view. 23 2. Relative Glare: The discomfort is happening because of the contrast of different luminance levels within the view although there is no absolute high luminance light source in the view. By effect/consequence: 1. Discomfort Glare: The source of glare may distract the observer even if performance remains unaffected (J. R. Coaton 1997, P30). 2. Disablity Glare: Excessively bright light sources produce disturbances and result in reducing visual performance and visibility, like a halo that impairs the visibility of nearby objects. It often happens along with discomfort. 3. Blinding Glare: Glare is so intense that for an appreciable period of time, eyes are incapable of sight (John E. Kaufman 1978, 1-4). By sources: 1. Direct Glare: Glare caused by high luminance or insufficiently shielded light sources within the field of vision. 2. Reflected Glare/ Veiling Glare: Glare as the result of a surface in the vision field which is reflecting a light source within or beyond the field of view. See Figure 20 (Stein 2006, P490). 24 Figure 20: Direct Glare and Reflected Glare By position: 1. Interior: Glare occurring within the room which typically refers to the glare caused by artificial lighting or daylighting from the windows or skylights. 2. Exterior: Glare occurring outside of buildings such as bright car headlights in a dark street, which may cause traffic problems or environmental impacts on surroundings or glare from the sun or reflected from specular surfaces. 1.5 Glare Analysis Methods 1.5.1 Psychological Statistical Survey Since glare is a visual sensation of human eyes, estimating the magnitude of glare is only possible when human subjects are involved to provide characterizations and assessments 25 of their experience. There are two methods of subjective psychological statistic survey types: one consists of inviting a group of subjects into a fully controlled test environment to execute some particular visual tasks then to answer questionnaires of their glare experience in the test environment; the more popular method is called relative visual performance (RVP). (Stein 2006, P499) This is a test of the subjects’ accomplishment of tasks in regard to speed and accuracy, representing the effectiveness of visual performance in a controlled/ given visual environment. Psychological statistical surveys evaluating discomfort glare severity in a specific situation are thought to be most accurate if they are organized and conducted in carefully controlled ways. However, they have certain limitations: can only be carried out in a built up environment, they are relatively time consuming and they require a relatively large survey sample covering possible variables such as age, gender, eye health, culture, etc. 1.5.2 Visual Comfort Probability Visual Comfort Probability (VCP) is a criterion established by the Illuminating Engineering Society of North America (IESNA). It is defined as the percentage of observers (of normal visual ability) who will report that they are comfortable in a specific environment which contains a number of individual glare sources in the interior space. IESNA published a set of standard conditions for which VCP could be calculated based 26 on 1000-lux illuminance, representative room dimensions, fixture positions and observers’ positions and view field (Stein 2006, P491). 1.5.3 The Luminance Histogram Method The Luminance Histogram Method was developed by Marc Schiler. It can be applied to predict possible discomfort glare in an existing scene which reflects human eyes’ experience of glare. It uses a common compact digital camera to take a photo of the studied scene. The resultant image is processed to produce a luminance histogram. If the histogram displays a pattern where most of the pixels are at a low level, but there is a substantial number of pixels at a high level, the high value is compared to the median of the low value. Current experiments show that a ratio of 3:1 or greater is coincident with the experience of glare. One practical method for this process uses the Adobe Photoshop program to generate a luminance map of the photo/ scene - changing the image mode into Grayscale. By applying the saved .Raw file of the image to a software called Rascal, the intensity of pixels data information of the image is extracted from the image file. The intensity of each pixel represents the luminance reading of each spot in the real scene. Then the intensity of pixel data file exported from Rascal is imported to another software called CULPLITE. CULPLITE is basically an Excel file. The intensity of pixel data is rearranged in the Excel file per its physical position in the image. A histogram is 27 generated based on the luminance intensity data. The x-axis is Intensity of Pixels which is representing the luminance data in the real world scene; the y-axis is the frequency (number of pixels of a specific intensity value within the whole image). By studying the bell shape and spike of histogram as well as the image in Photoshop, the observer records the high and low pixel values for both bell and spike regions. The ratio of median value of the spike region to the median value of the bell region is called Schiler Index. Compare Schiler Index to the Schiler histogram criteria to make a prediction of glare possibility. If Schiler Index is larger than 3:1, then there is a high probability of glare experience; if Schiler Index is between 2:1 and 3:1, then it has a relatively low probability of glare problem in scene; if Schiler Index is below 2:1, then there is minimum possibility of glare problem. The 3:1 ratio is an approximate empirical value derived from an initial survey of individuals in test environments. (Schiler 2000) 1.5.4 Calculated Glare Indices Glare indices are the numbers describing the subjective magnitude of glare discomfort which are derived from a range of experimental and survey of subjective glare sensation studies. These experiments were conducted with artificial light sources. The higher the index value the worse the glare sensation. The major factors which are studied as contributing to glare indices calculations are glare source luminance, the general field 28 luminance, the solid angle subtended by the source, the geometrical relationship between glare source and observer’s sight line, and the general field luminance. The subjective sensation of glare is generally related to the numerical quantity. (Jan Wienold 2006) G: Subjective Sensation; e, f and g: are weighting exponents; typically f = e/2; : is a function of the displacement angle which is the angle displacement of the source from observer’s line of sight; it describes the relative importance of glare sources as they occur off the line of sight. : Glare source luminance, cd/m² ; : General field of luminance which also referred as background luminance, is the control parameter of the adaption vision luminance level, cd/m² : The solid angle subtended by the source, steradians (sr).) 29 There are several currently popular glare indices which are derived from the quantity G above. They all based on empirical mathematical calibration of indices number towards glare sensation survey (Jan Wienold 2006). They are: 1. Unified Glare Rating(UGR) UGR= p: the Guth position index is a function of angular displacement of the source from the direct line of sight and increases with increasing angular displacement. n: The number of glare sources. This is an equation developed by International Commission on Illumination which is also known as CIE (by French Commission Internationale de l´ Eclairage). It restricted to glare sources which are within the relative small solid angle of 3× 10 -4 to 10 -1 . (Jan Wienold 2006) Building Research Station Formula (BRS) BRS= This is an equation developed by Petherbridge and Hopkinson at the Building Research Station in England (Jan Wienold 2006). 30 The relationship between UGR, BRS and typical glare criterion is as follow, Table 1 (Iwata n.d.).and Table 2 (Melo n.d.) Glare Criterion BRS/BGI Imperceptible 10 Just Perceptible 13 Perceptible 16 Just Acceptable 19 Uncomfortable 22 Just Intolerable 25 Intolerable 28 Table 1: Typical Glare Criterion and BRS Glare Criterion UGR Just Perceptible 10~13 Just Acceptable 15~17 Just Uncomfortable 22~24 Just Intolerable 27~30 Table 2: Typical Glare Criterion and UGR 31 Chapter 2: Glare Analysis Out of the four glare analyses mentioned in the previous chapter, two out of them: the Luminance Histogram Method and Calculated Glare Indices require no people tests directly involved during glare analysis. They are based on the luminance and/or illuminance data and combine human visual sensation factors into certain equations to generate the glare probability and severity levels. This numerical research methodology makes these two methods very valuable and convenient in glare analysis and prediction. Therefore this thesis is studying the Luminance Histogram Method and Calculated Glare Indices. Luminance map is the key factor for either method. For glare analysis of an existing study space, luminance data could be collected by using luminance meter, however, this process is very tedious and time consuming. Research indicates that photographs can be used as luminance collecting tools, for example CCD camera. Due to its expensive price, researchers try to use other less expensive digital camera (Luminance Histogram Method uses compact digital camera). And recently, High Dynamic Range Images (HDR/ HDRI) is getting more popular and considered as a better way to record luminance data than traditional photographs. Therefore this thesis is focused on the use of HDR images to generate the luminance map. For glare prediction of a not-yet-constructed 32 space, several computer modeling softwares are capable of generating luminance and illuminance maps for glare analysis. 2.1 High Dynamic Range Image High Dynamic Range image (HDR) is a type of processed digital image that contains much higher dynamic range information in aspects like luminance and color beyond the raw picture capability of the digital camera. The wider extent dynamic range image is capable of more exactly representing the differential light levels between the brightest area (over exposed in a single exposure image) and darkest area (under exposed in a single exposure image). 2.1.1 Digital Camera and Camera Response Curve Just as film cameras, digital cameras capture scenes by capturing incoming light when the camera aperture opens. The light ray sensor in a digital camera is an electronic chip. It turns the light energy which strikes the sensor into electrical signals. There are two types of sensors commonly used in digital cameras, Charge Coupled Device (CCD), Figure 21 (Wiki, Camera CCD Sensor n.d.), which is integrated with a photoelectric sensor and Complementary Metal-oxide-semiconductor Active pixel sensor. 33 The amount of detail in each picture taken by a digital camera is called the resolution. Resolution is measured in pixels. The larger the number of pixels the more information is recorded from the sensor. The size of sensor and size of lens are influential elements for a broader definition of resolution. Figure 21: CCD Sensor The sensor can only record the number of photons striking it when the aperture is open, in other words proportional to the luminance of the scene with the factor of proportionality depending on the size of the aperture and the period of opening. The color image is made by applying color filters in the digital camera. There are several types of filter invented and used by different digital camera manufacturers. Among them, the most popular type is called the Bayer Filter, Figure 22 (Wiki, Bayer Arrangement of Color Filter n.d.). The Bayer filter is a mosaic of color filters set in front of a square and 34 gridded photo-sensor. Bayer filters are also divided into several categories by different color combinations: GRGB (double green, red, blue), CMYW (cyan, magenta, yellow, white), and RGBW (red, green, blue, white). Since human eyes have different color sensitivity among the colors, the Bayer filter is designed to as much as possible mimic people’s visual experience. For example, human eyes are more sensitive to green than to red and blue, in GRGB filter; the ratio of green area to blue area and red area is 2:1:1. Cameras using Bayer filters have a series of de-mosaicing algorithms in order to interpolate green, blue and red for each pixel. The Camera company-- Foveon invented another type of filter called the vertical color filter, Figure 23 (Wiki, Foveon X3 Sensor n.d.). The color filter is actually the sensor itself. The vertical color filter takes advantage of certain material’s wavelength-dependent absorption characteristics. The sensor is made up with three layers of different wavelength-dependent absorption materials to detect and separate color composition in incoming light rays. A third recently invented type of color filter uses a refracting microlens to separate light onto specific photoreceptors, as Figure 24 (Wiki, Nikon's Dichroic Color Seperation Patent Drawing n.d.). These photoreceptors are designed to only be capable of detecting specific wavelength of light, like green, blue and red. As claimed by Nikon, the inventor, this type of filter can theoretically provide the best quality color images. 35 Figure 22: Bayer Arrangement of Color Filters Figure 23: Foveon X3 Sensor Figure 24: Nikon's Dichroic Color Seperation Patent Drawing To typical film cameras, lenses are usually divided into 4 categories based on the focal length: Standard Lenses (single lens ~50mm) for a 35mm receptor area providing an 36 angle of view close to human near-field vision; zoom lenses (35-100mm); Wide Angle Lenses (<50mm); Medium Telephoto Lenses (85-135mm); and Long Telephoto Lenses (>135mm). There are other special function lenses: Fisheye Lenses (7-16mm), which by distorting the perspective, can create a circular 180-degree view field which is theoretically close to the total human eye’s view field. For digital cameras, because the photon sensor chips for each model of camera varies the traditional lenses division criteria may vary. For typical compact cameras, the sensor chips are much smaller than film frame (sensitive area about 36 x 24mm), a standard lens (single lens ~50mm) would be a Long Telephoto lens since the effective focus length would be about 200mm. For compact digital cameras lenses are not inter changeable, the lens data showed on spec-sheet are the relative focus length which are calculated values. The calculation based on the sensor (CCD) size compared with film sensitive area size. They are called relative/ effective focus length. Exposure is determined by camera ISO speed, exposure time, aperture size. Exposure = log2 ( Aperture2 * (1/Shutter speed) * (ISO Speed/100) ) (Jacobs, High Dynamic Range Imaging and its Application in Building Research 2007) 37 The Camera Response Curve is a polynomial function that models the accumulated radiometric non-linearities of the image acquisition process (such as gamma correction, A/D conversion, image digitizing, and various mappings, without addressing the individual source of each non-linearity) (Inanici 2004). The response curve of each photo-sensitive chip varies even within the same brand, model or make although the variation may not necessarily be significant. For some popular camera types, the response curves have been generated and stored in a software named Photosphere (which will be discussed in Chapter 3). On the other hand, Photosphere also provides an option of creating new response curves using a sequence of bracketed multiple exposure images for a static scene. 2.1.2 High Dynamic Range Image Software & Format High Dynamic Range Image Formats: Log Forma: was invented and developed by Pixar in their film recording. It is a 33-bit/pixel log encoding for RGB value. It is implemented in Sam Leffler's TIFF library. Over 3.5 orders of magnitude have been already encoded, with a relative accuracy of 0.4% (Jacobs, WebHDR n.d.). 38 RGBE Format: was first launched in 1987 in Radiance. It uses a 32-bit/pixel floating point format. Over 76 orders of magnitude have been already encoded, with a relative accuracy of 1% (Jacobs, WebHDR n.d.). LogLuv Format: is a 32-bit/pixel logLuv encoding collected in Sam Leffler's TIFF library. It has the capability of covering the full range of perceivable colors even within imperceptible levels. The luminance range is beyond38 orders of magnitude, but with a relative accuracy of 0.3% (Jacobs, WebHDR n.d.). OpenEXR : was developed to expand the dynamic range and increase color precision of images for computers. It is an Open Source file structure (Jacobs, WebHDR n.d.). HDR making Software: JPEG-HDR :The JPEG-HDR creates very small files sizes (Jacobs, WebHDR n.d.). Hdrgen: This is the HDR composition engine applied in WebHDR and Photosphere. The process automatically aligns the bracket images to compensate for possible camera shift or movement (Jacobs, WebHDR n.d.). Formats: RGBE, LogLuv TIFF, OpenEXR Bracket: This is a cross-platform software for HDR images. Bracket is capable of creating, viewing, analyzing, calibrating, tone-mapping, and displaying HDR images. It is 39 a software more geared towards artistic photography. It is also capable of accomplishing some basic image editing technics like image cropping, rotating, and resizing (Jacobs, WebHDR n.d.). Formats: RGBE, LogLuv TIFF, FP TIFF, OpenEXR Picturenaut: This is a German software. It provides an interesting interface to generate and manipulate HDR images. Basically, it is a software intended for photography (Jacobs, WebHDR n.d.) Formats: RGBE, LogLuv TIFF, FP TIFF, OpenEXR, PFM Photosphere: (formerly called Photophile.) This is an image cataloging and browsing software which also provides HDR images creation as well as camera response curve generation and storage from both automatic and hand-held exposure bracketed images. It also provides the luminance histogram for the created HDR image. It automatically aligns the sequence of images in order to prevent inaccuracy caused by camera movement or shifting. Although it allows basic image refining processes like tone-mapping, it is more intended as scientific software than artistic photographic software (Jacobs, WebHDR n.d.). Formats:RGBE, LogLuv TIFF, 48bit TIFF, FP TIFF, OpenEXR 40 2.2 HDR analysis software 2.2.1 Radiance Radiance is a suite of programs which can be used for rendering/ simulation and lighting analysis. For rendering, the scene geometry, luminaires, time, date materials, and sky conditions are input information to set up a scene. Rendering output are color images, contour plots and numerical values. Analysis can generate lighting information such as spectra. This is a script program under UNIX. Radiance has no limitations on the input geometry, the materials and luminaires that may be simulated. Radiance as a simulation and calculating engine has been plugged into or integrated with many programs and software such as Daysim, Evalglare, 3D Max Design, Radiance Desktop, and Ecotect (Radiance n.d.). These programs as well as software are broadly used by architects and engineers in illumination prediction, visual quality and appearance rendering; and by researchers to do daylighting analysis and other lighting studies such as glare and illuminant studies. 2.2.2 GlareIndices (C++ Program) This is a simple program written for glare analysis of RGBE formatted HDR images analysis. It can be run on the Windows operating system. The goal of writing this program was to do simple and exclusive glare analysis. The analysis output is a series of 41 Glare Indices and Schiler Histogram Index. The basic code for reading HDR image luminance information is referenced from www.graphics.cornell.edu: f = fopen (image_filename, "rb"); RGBE_ReadHeader (f, & image_width, & image_height, NULL); image = (float*) malloc (sizeof (float)*3*image_width* image_height); RGBE_ReadPixels_RLE (f, image, image_width, image_height); fclose(f). This code was written by Bruce Walter and based on Radiance. It is an open source, program that is only capable of reading non-compressed hdr format HDR images. The camera parameters like focal length and sensor size need to be manually set in, which means, they need to be record by the users and typed into the software when analyzing each HDR image. One good source for reading focal length is using a photography process software called Lightroom, it shows the actual focal length rather than the calculated focal length some other software may provide. The sensor size is the characteristic of the camera which could be read from camera menu or statement.The Glare searching algorithm scans each pixel’s luminance values from upper left to bottom right, one by one and line by line. Same as Radiance (Ward 1992), a pixel is defined as a glare pixel when the luminance value is higher than 7 times of the average luminance value within the image. Once a glare pixel is found, the scan pattern changes to search the adjacent (upper, below, left and right) pixels for glare pixels and connect them into one glare zone. After reaching the boundary of that glare zone the normal scan pattern resumes and continues until a new 42 glare pixel is found which is not part of any previously defined glare zones. By repeating this scanning and searching process, the program finds and defines the glare zones for the entire image. The definition of the position of glare zone is the glare pixels’ coordinates X and Y axes, with the (0, 0) pixel on the bottom left of the image. The center of the glare zone (Xcenter, Ycenter) indicates the position of the glare zone, which is defined by the equations below: Xcenter=1/2 (Xleft+ Xright); Ycenter=1/2(Ytop+Ybottom). The solid angle of glare source (glare zone) is the area (number of pixels) divided by the square of the focal length of the camera. The equations for Glare Indices are calculated by the equations mentioned in the previous chapter and the luminance data, position index, and solid angle are collected by the hrd. Image. The histogram for hdr image is natural logarithm processed for its luminance axis (X axis) to provide a better and clear view of the luminance range curve. By looking at the high dynamic range histogram, the user can then define the “bell” shape and “spike” shape for the Luminance Histogram Method by typing in the maximum and minimum value to define the range for both “bell” and “spike”. The program will automatically calculate the 43 average luminance values for both “bell” and “spike” shape and the ratio between them and then gives the Schiler Histogram Rating of glare probability. Interface is as following Figure 25. Figure 25: GlareIndices (C++ Program) Interface The operation steps: 1. Open the HDR image (in Radiance RGBE format, .hdr) on the computer; 2. Name and specify a folder to save the Histogram (.csv); 3. Input the basic digital camera parameters: Sensor Size and Focal Length; 4. Hit the “Process 1” button to calculate BGI and UGR and histogram; 44 5. When the program finish the generation of histogram a warning dialogue pops out, Figure 26. 6. Open the histogram file with Excel and use the “Scatter Diagram” function in Excel to generate the Histogram. Define the Bell Shape and Spike Shape. 7. Input the luminance values for the defined Bell Shape maximum and minimum luminance: BellMax, BellMin as well as the defined Spike Shape maximum luminance and minimum luminance: SpikeMax and SpikeMin. Figure 27 8. Hit the “Process 2” button. The program will calculate the Schiler Index and the conclusion. Figure 28. Figure 26: Warning Dialog in Step 4 45 Figure 27: Input BellMax, BellMin, SpikeMax and SpikeMin in Step 7 Figure 28: A Demonstration of Final Result Output 46 Chapter 3: HDR Experiments The first and basic experiment is to test the reliability of HDR images as a luminance mapping tool. As described in the previous chapter, by taking a series of exposure bracketed (eg. -2EV, -1EV, 0EV, +1EV, +2EV) low dynamic range photos and processing them by certain software, high dynamic range (HDR) images are created and stored in typically 16-bit image formats. As the luminance range for each HDR image is largely expended in image storage format and camera limitation got pushed by the way of taking photos, the luminance reading for each pixel in HDR image is theoretically more close to real world scene as compared with traditional low dynamic range image. The following experiment is going to demonstrate the accuracy of HDR image luminance map of controlled illuminated circumstances compared with luminance meter readings of the real world scene. With the experiment, the author is trying to find out the relationship between HDR image luminance mapping accuracy and cameras brands and models, lighting environment, the performance limitation of HDR images as a luminance mapping tool. Three experiments were carried out. 47 3.1 Experiment Test 1: Camera (with Photosphere) The test room is lit purely by daylight. The light level is controlled by two shading devices: blades and venetian blind system. By rolling up or down the blades and venetian blinds, diverse light levels were achieved in the room. Total seven light levels were presented in the experiment. Four diffused surfaces with different reflectance and respectively of four different colors: white, light gray, dark gray and black were fixed on the wall. On each surface, a square target was drawn. These targets are clustered together. So the illuminance level on each square could be more or less the same. Luminance readings for each surface were taken within these targets. Minolta Luminance Meter LS-110 with view angle of 1/3° was used in the experiment. The luminance meter was mounted on a fixed tripod to take luminance readings from the marked targets on the white, light gray, dark gray and black boards. Illuminances on the center of tested surface were taken by IALD 61-680 which represent the overall illuminance for the four targeted areas. Illuminance values were recorded. Figure 29. Two types of cameras were used in the experiments: Canon Rebel T2i is a single-lens Reflex (SLR) digital camera with an 18-135mm lens; Panasonic ZS1 is a compact digital 48 camera. In order to make sure the camera settings are not the main effective reasons for difference between these two cameras. The cameras have the settings as the following table. Canon Rebel T2i Panasonic ZS1 White balance Daylight Landscape (Daylight) Aperture A-DEP (5.6) Auto (4.5) File size 3456× 2304 2560× 1920 Focus Auto Auto Focal Length 120mm 29.9mm Sensor Size 22.3mm× 14.9mm 1/ 2.33 Table 3: Test 1 Cameras Settings Figure 29: Illuminance Meter Reading 49 The luminance readings for each measured surface (in the target) is as follows: Luminance Meter Readings Illuminance fc Black Surface cd/m2 Dark Gray Surface cd/m2 Light Gray Surface cd/m2 White Surface cd/m2 3.4 1.31 2.8 7.81 16.51 8.8 2.84 6.26 16.26 34.65 17.8 6.14 15.01 36.95 81.7 31 9.26 21.3 53.9 117 133.1 45.99 96.67 272.9 595.8 197.0 62.24 146.0 410.8 899.2 251.0 78.21 183.6 522.7 1135 Table 4: Luminance Meter Readings of Each Measured Block HDR and luminance mapping software: Photosphere Camera response curves for both cameras were generated by applying 5 photos with -2EV, -1EV, 0EV, +1EV and +2EV of an interior scene which contains very bright sky through the window and dark object inside the room. A series of exposure bracketed photos were taken to make the correct camera response curve for one particular camera. The captured scene contains direct sunlight, outdoor luminances (sky), shadows and relatively dark interior surfaces. The expanse of luminance range makes better correspond between camera response curve and real world luminance scene. 50 The process of making a camera response curve is demonstrated by Panasonic ZS1 as following. Five bracketed photos as following Figure 30~34 were taken. 51 Figure 30: Panasonic -2EV Exposure Figure 31: Panasonic -1EV Exposure Figure 32: Panasonic 0EV Exposure Figure 33: Panasonic +1EV Exposure Figure 34: Panasonic +2EV Exposure By uploading images Photosphere. Camera response curve are generated. 52 Panasonic ZS1 Response Curve: 2 1.10259 -0.150364 0.0477719 2 1.16048 -0.21209 0.0516112 2 1.1195 -0.163215 0.0437146 Red Channel: Luminance =1.10259x 2 -0.150364x+0.0477719 Green Channel: Luminance =1.16048 x 2 -0.21209x+0.0516112 Blue Channel: Luminance=1.1195x 2 -0.163215x+0.0437146 The HDR image created by Photosphere: Figure 35: Photosphere HDR Image 53 With the saved camera response curve, the photos of experiment scenes previously described were processed to HDR images in Photosphere. In Photosphere, by selecting the target area on the photo, luminance values are calculated (within Photosphere) and recorded. But since computer monitor is not able to provide the extended dynamic range, this high dynamic ranged image has been tone mapped to make it closer to human visual experience. Five photos were taken under a same illuminance with -2EV, -1EV, 0EV, +1EV, +2EV. (eg. the following photos Figure 36~40, were taken by Panasonic ZS1 when the illumiance was 31 fc.) 54 Figure 36: Panasonic -2EV Exposure Figure 37: Panasonic -1EV Exposure Figure 38: Panasonic 0EV Exposure Figure 39: Panasonic -+1EV Exposure Figure 40: Panasonic +2EV Exposure 55 These five photos were processed in Photosphere to generate an HDR image as following, Figure 41. But since computer monitor is not able to provide the extended dynamic range, this high dynamic ranged image has been tone mapped to make it closer to human visual experience. Histogram for luminance over the whole image is generated based on original luminance value (before tone mapping). And the luminance information contained in the .hdr file is the real luminance which is the luminance before tone mapping. Since the expanded luminance range and precision over the range, histogram is a logarithm graph across the X axis (which is the luminance).. Figure 41: Test 1 Panasonic Photosphere HDR Image 56 The luminance readings for the four targeted areas of the four surfaces from Canon Rebel T2i Photosphere HDR are listed in the following tables. Canon HDR Illuminance fc Black Surface cd/m2 Dark Gray Surface cd/m2 Light Gray Surface cd/m2 White Surface cd/m2 3.4 0.912 2.98 8.94 19.1 8.8 2.16 6.22 17 34.9 17.8 4.08 13.3 29.4 69.7 31 6.92 22 58.1 129 133.1 33.8 101 248 578 197 51.2 146 362 778 251 66.3 198 503 1190 Table 5: Canon Rebel T2i HDR Image Luminance Readings from Photosphere The comparison between HDR image luminance reading and luminance meter reading for each target was addressed under different light levels. The comparison contains three aspects: 1. The ratios of luminance three colored surfaces over white surface (the reflectance ratios); 2.The difference of absolute luminance reading between HDR image luminance readings and 3. The magnitude of the difference, defined as error. Error= 57 Since the four surfaces in the scene have fixed reflectance values, the ratios between Black/White, Dark Gray/White, Light Gray/White should be consistent over the different illuminance levels. For the luminance meter readings: Illuminance Black/White Dark Gray/White Light Gray/White White/White 3.4 0.079345851 0.169594185 0.473046638 1 8.8 0.081962482 0.180663781 0.469264069 1 17.8 0.075152999 0.18372093 0.452264382 1 31 0.079145299 0.182051282 0.460683761 1 133.1 0.077190332 0.162252434 0.458039611 1 197 0.069217082 0.162366548 0.456850534 1 251 0.068907489 0.161762115 0.460528634 1 Average 0.075845933 0.171773039 0.461525376 1 Standard Deviation 0.005081316 0.010095623 0.007235323 0 Table 6: Luminance Ratios (over White Surface) from Luminance Meter 58 The Canon T2i Photosphere HDR image ratios: Illuminance Black/White Dark Gray/White Light Gray/White White/White 3.4 0.047749 0.15602094 0.46806283 1 8.8 0.061891 0.1782235 0.48710602 1 17.8 0.058537 0.19081779 0.42180775 1 31 0.053643 0.17054264 0.4503876 1 133.1 0.058478 0.17474048 0.42906574 1 197 0.06581 0.18766067 0.46529563 1 251 0.055714 0.16638655 0.42268908 1 Average 0.057403 0.17491322 0.44920209 1 Standard Deviation 0.005823 0.01207728 0.02553333 0 Table 7: Luminance Ratios (over White Surface) from Canon T2i HDR 59 The Panasonic ZS1 Photosphere HDR image ratios: Illuminance Black/White Dark Gray/White Light Gray/White White/White 3.4 0.039077 0.081026 0.402051 1 8.8 0.035809 0.1313 0.490716 1 17.8 0.033901 0.113003 0.441176 1 31 0.031417 0.12 0.485 1 133.1 0.064578 0.198093 0.495913 1 197 0.057725 0.162479 0.485569 1 251 0.056555 0.161954 0.479434 1 Average 0.04558 0.138265 0.468551 1 Standard Deviation 0.013562 0.03878 0.034376 0 Table 8: Luminance Ratios (over White Surface) from Panasonic ZS1 HDR By studying on the average ratios, it is found that the Canon T2i Photosphere HDR image luminance readings for dark gray surface and light gray surface over white surface are very close to luminance meter readings, but for the ratio of black surface over white surface is lower than the standard luminance meter readings. Panasonic ZS1 Photosphere HDR images have very close ratios of light gray surface luminance over white surface luminance to real world luminance meter reading ratio. But for the black surface and dark gray surfaces, the ratios are lower than luminance meter reading ratios. 60 By studying the standard deviation of the ratios, it is easy to find that both cameras’ HDR luminance readings ratios are not as stable as luminance meter readings ratio. However, Canon T2i has better performance than Panasonic ZS1. And the standard deviation of the ratio between light gray surface and white surface is larger than the ratios of black surface/white surface and dark gray surface/white surface, but still in acceptable range. This is unexpected. The comparison of the absolute luminance readings from Canon T2i HDR image luminance readings and luminance meter of the real world scene readings of the four measured surfaces under tested light levels are plotted in Figure 46. Figure 42: Overall Comparison between Canon T2i HDR and Luminance Meter 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 Canon T2i Photosphere HDR Luminance Reading cd/m2 Standard Luminance Reading cd/m2 (Luminance Meter) 61 Then the comparison is divided based on the measured surfaces: White Surface Luminance Meter cd/m2 Canon HDR Luminance cd/m2 16.51 19.1 34.65 34.9 81.7 69.7 117 129 595.8 578 899.2 778 1135 1190 Table 9: Comparison of White Surface Canon T2i HDR and Luminance Meter Figure 43: Comparison of White Surface Canon T2i HDR and Luminance Meter 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 Canon HDR Image Luminance cd/m2 Standard Luminance cd/m2 (Luminance Meter) 62 Light Gray Surface Luminance Meter cd/m2 Canon HDR Luminance cd/m2 7.81 8.94 16.26 17 36.95 29.4 53.9 58.1 272.9 248 410.8 362 522.7 503 Table 10: Comparison of Light Gray Surface Canon T2i HDR and Luminance Meter Figure 44: Comparison of Light Gray Surface Canon T2i HDR and Luminance Meter 0 100 200 300 400 500 600 0 100 200 300 400 500 600 Canon HDR Image Luminance cd/m2 Standard Luminance cd/m2 (Luminance Meter) 63 Dark Gray Surface Luminance Meter cd/m2 Canon HDR Luminance cd/m2 2.8 2.98 6.26 6.22 15.01 13.3 21.3 22 96.67 101 146 146 183.6 198 Table 11: Comparison of Dark Gray Surface Canon T2i HDR and Luminance Meter Figure 45: Comparison of Dark Gray Surface Canon T2i HDR and Luminance Meter 0 50 100 150 200 0 50 100 150 200 Canon HDR Image Luminance cd/m2 Standard Luminance cd/m2 (Luminance Meter) 64 Black Surface Luminance Meter cd/m2 Canon HDR Luminance cd/m2 1.31 0.912 2.84 2.16 6.14 4.08 9.26 6.92 45.99 33.8 62.24 51.2 78.21 66.3 Table 12: Comparison of Black Surface Canon T2i HDR and Luminance Meter Figure 46: Comparison of Black Surface Canon T2i HDR and Luminance Meter 0 20 40 60 80 0 20 40 60 80 Canon HDR Image Luminance cd/m2 Standard Luminance cd/m2 (Luminance Meter) 65 The error for the measured four surfaces could be calculated, as Table 13 shows. Canon Rebel T2i Illuminance fc Black Surface % Dark Gray Surface % Light Gray Surface % White Surface % 3.4 -30.38% 6.43% 14.47% 15.69% 8.8 -23.94% -0.64% 4.55% 0.72% 17.8 -33.55% -11.39% -20.43% -14.69% 31 -25.27% 3.29% 7.79% 10.26% 133.1 -26.51% 4.48% -9.12% -2.99% 197 -17.74% 0.00% -11.88% -13.48% 251 -15.23% 7.84% -3.77% 4.85% Table 13: The Errors of Measured Surfaces under Different Light Levels The overall error for the tested scenarios is plotted as follow Figure 47 and Figure 48. As it is indicated from the chart, for black surface, the Canon T2i HDR luminance readings are always lower than luminance meter and for the rest of the cases: dark gray surface, light gray surface and white surface the Canon T2i HDR luminance readings errors are scattered above or below zero. For extreme situations, the lower the reflectance is, the reading is relatively lower than luminance meter, and on the other hand, the higher the reflectance the reading is higher than the luminance meter reading. Therefore, the surface reflectance is one element that is affecting the accuracy of luminance reading from HDR images. The luminance error is independent to different illuminance levels. In other 66 words, HDR images as a luminance mapping tool, the performance is consistent over various illuminance conditions. For most of the tested circumstances (except black surface), the luminance reading errors of Canon T2i HDR images are among -20% to 20% and for dark gray surface and light gray surface and white surface, most of the errors are among -10% to 10%. Canon T2i HDR luminance readings are reliable. However, there is a noticeable set of luminance values behaves differently than the rest of test sets: the illuminance 17.8fc. The errors for the four surfaces (four measured areas) are all negative. This could be caused by changing of light environment between luminance recordings (taking photos and use luminance meter. Also another noticeable set is when the illuminance is 3.4 fc. When the light level is extremely low as 3.4 fc, the Canon T2i Photosphere HDR image luminance readings are relatively higher than the other light levels. Low reflectance surface black surface has a much lower reading in HDR image than the real world luminance but with the same scene the other higher reflectance surfaces have much higher luminance readings in HDR image than the luminance meter readings. It seems that HDR image exaggerate the error tendency when the light level is extremely low. The camera’s internal tone mapping may increase the high reflectance surface luminance (over expose) and under expose the low reflectance surface luminance. This indicates that accuracy of typical 5-stops bracketed exposure HDR image as a luminance mapping tool may have light level limitations. 67 Figure 47: Canon T2i HDR Luminance Error ( X axis is logarithm, log2) Figure 48: Canon T2i HDR Luminance Errors -40.00% -30.00% -20.00% -10.00% 0.00% 10.00% 20.00% 1 4 16 64 256 1024 Error Standard Luminance cd/m2 (Luminance Meter) Canon T2i Photosphere HDR Image Lumiance Errors Black Surface Dark Gray Light Gray White Surface -40.00% -30.00% -20.00% -10.00% 0.00% 10.00% 20.00% 0 50 100 150 200 250 300 Error Illuminance fc Canon T2i Photosphere HDR Image Luminance Errors Black Surface Dark Gray Surface Light Gray Surface White Surface 68 Panasonic ZS1 under same experiment condition: Panasonic HDR Illuminance fc Black Surface cd/m2 Dark Gray Surface cd/m2 Light Gray Surface cd/m2 White Surface cd/m2 3.4 0.762 1.58 7.84 19.5 8.8 1.35 4.95 18.5 37.7 17.8 2.19 7.3 28.5 64.6 31 3.77 14.4 58.2 120 133.1 23.7 72.7 182 367 197 34 95.7 286 589 251 44 126 373 778 Table 14: Panasonic HDR Image Luminance Readings from Photosphere The comparison of the luminance readings from Panasonic ZS1 HDR image luminance readings and luminance meter of the real world scene readings of the four measured surfaces under tested light levels are plotted in Figure 49. 69 Figure 49: Overall Comparison between Panasonic HDR and Luminance Meter 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 Panasonic ZS1 Photosphere HDR Image Lumiance cd/m2 Standard Luminance Reading cd/m2 (Luminance Meter) 70 Then the comparison is divided based on the measured surfaces: White Surface Lumiannce Meter cd/m2 Panasonic HDR Luminance cd/m2 16.51 19.5 34.65 37.7 81.7 64.6 117 120 595.8 367 899.2 589 1135 778 Table 15: Comparison of White Surface Panasonic ZS1 HDR and Luminance Meter Figure 50: Comparison of White Surface Panasonic HDR and Luminance Meter 0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 Panasonic HDR Image Luminance cd/m2 Standard Luminance cd/m2 (Luminance Meter) 71 Light Gray Surface Lumiannce Meter cd/m2 Panasonic HDR Luminance cd/m2 7.81 7.84 16.26 18.5 36.95 28.5 53.9 58.2 272.9 182 410.8 286 522.7 373 Table 16: Comparison of Light Gray Surface Panasonic HDR and Luminance Meter Figure 51: Comparison of Light Gray Surface Panasonic HDR and Luminance Meter 0 100 200 300 400 500 600 0 100 200 300 400 500 600 Panasonic HDR Image Luminance cd/m2 Standard Luminance cd/m2 (Luminance Meter) 72 Dark Gray Surface Lumiannce Meter cd/m2 Panasonic HDR Luminance cd/m2 2.8 1.58 6.26 4.95 15.01 7.3 21.3 14.4 96.67 72.7 146 95.7 183.6 126 Table 17: Comparison of Dark Gray surface Panasonic HDR and Luminance Meter Figure 52: Comparison of Dark Gray Surface Panasonic HDR and Luminance Meter 0 50 100 150 200 0 50 100 150 200 Panasonic HDR Image Luminance cd/m2 Standard Luminance cd/m2 (Luminance Meter) 73 Black Surface Lumiannce Meter cd/m2 Panasonic HDR Luminance cd/m2 1.31 0.762 2.84 1.35 6.14 2.19 9.26 3.77 45.99 23.7 62.24 34 78.21 44 Table 18: Comparison of Black Surface Panasonic HDR and Luminance Meter Figure 53: Comparison of Black Surface Panasonic HDR and Luminance Meter 0 20 40 60 80 100 0 20 40 60 80 100 Panasonic HDR Image Luminance cd/m2 Standard Luminance cd/m2 (Luminance Meter) 74 The error of Panasonic ZS1 luminance readings are calculated as Table 19. Illuminance fc Black Surface % Dark Gray Surface % Light Gray Surface % White Surface % 3.4 -41.83% -43.57% 0.38% 18.11% 8.8 -52.46% -20.93% 13.78% 8.80% 17.8 -64.33% -51.37% -22.87% -20.93% 31 -59.29% -32.39% 7.98% 2.56% 133.1 -48.47% -24.80% -33.31% -38.40% 197 -45.37% -34.45% -30.38% -34.50% 251 -43.74% -31.37% -28.64% -31.45% Table 19: Panasonic Luminance Reading Errors under Different Light Levels Panasonic ZS1 HDR luminance readings for black surface are much lower than the luminance meter readings. The error range for black surface is from -41.83% to -64.33%. The error range for dark gray surface is from -20.93% to-51.37%; light gray surface error range is from 13.78% to -33.31%; White surface error range is from -38.40% to 18.11%. The overall luminance reading error for ZS1 HDR is in the negative side. Panasonic ZS1 HDR image as a luminance mapping tends to give lower readings than the luminance meter. The errors are not consistent or proportional to the illuminance levels on the test surfaces. Same as in Canon T2i HDR image, when the light level is 17.8 fc, the luminance readings from HDR image of the four measured surfaces are relatively lower than the other sets. However, for light gray surface and white surface, the luminance 75 readings of Panasonic ZS1 Photosphere are tend to be lower than the luminance meter readings when the light level is high, for example when light levels are 133.1fc, 197fc and 251fc. The overall luminance reading errors in Panasonic HDR images tend to be negative. The overall error charts for Panasonic ZS1 are as follow: Figure 54: Panasonic HDR Luminance Error ( X axis is logarithm, log 2) -70.00% -60.00% -50.00% -40.00% -30.00% -20.00% -10.00% 0.00% 10.00% 20.00% 30.00% 1 4 16 64 256 1024 Error % Standard Luminance cd/m2 (Luminance Meter) Panasonic ZS1 Photosphere HDR Image Luminance Errors Black Surface Dark Gray Surface Light Gray Surface White Surface 76 Figure 55: Panasonic HDR Luminance Error From the analysis of Test 1 results, three conclusions could be drawn: 1. Camera model is an important affecting element for HDR images luminance reading; 2. Reflectances of measured surface which are captured by the photo are affecting the accuracy of luminance reading for these surfaces; 3. For the tested range of illuminances on the surfaces, Canon T2i HDR images as a luminance mapping tool, works consistently under different light levels. But for Panasonic ZS1 HDR images, it is found that the light level may affect the accuracy of high reflectance surfaces luminance reading. -70.00% -60.00% -50.00% -40.00% -30.00% -20.00% -10.00% 0.00% 10.00% 20.00% 30.00% 0 50 100 150 200 250 300 Error Illuminance fc Panasonic ZS1 Photosphere HDR Image Luminance Error Black Surface % Dark Gray Surface % Light Gray Surface % White Surface % 77 By comparing Canon Rebel T2i HDR images and Panasonic ZS1 images, the performance of Canon Rebel T2i is better than Panasonic ZS1. As described earlier, Canon Rebel T2i as a digital SLR camera has a 22.3× 14.9 mm sensor (CMOS) and Panasonic ZS1 as a compact digital camera has a 1/ 2.33” (10.9 mm diagonal) (CCD). The size of photonic sensor could be one reason to explain the better performance of Canon Rebel T2i digital camera. 3.2 Experiment Test 2: Reflectance As found from Test 1, reflectance of test (luminance recording) surface is affecting the accuracy of Canon Rebel T2i HDR image as a luminance mapping tool. So a second test was conducted to test the influence tendency of reflectance to Canon Rebel T2i HDR luminance reading accuracy. The experiment setting is similar to Test 1: a daylight lit test room. The light level is controlled by shades and a venetian blind system. By rolling up or down the shades and adjusting the venetian blinds, various light levels were achieved in the room. Five light levels were presented in the experiment. A Minolta Luminance Meter LS-110 with view angle of 1/3° was used in the experiment. The luminance meter was mounted on a fixed tripod. 78 Gretag MacBeth Color Checker, Figure 56 is hung on the vertical wall with a light grey surface as the background. The color checker is under diffused daylight. The last row of blocks on the checker is the target row. It was placed in the center of the camera view to minimize the vignette effect. The information is as the following Table.20. Since each block is a slightly under 2 inches square and the distance between the checker and luminance meter (camera) is about 6 feet, the light level on each block is quite even. Luminance meter aims at the center of each block. The measured angle of the meter is well within the edges of each block. Figure 56: Gretag MacBeth Color Checker 79 The basic information of Gretage MacBeth Checker: Block Index Description Munsell Notation CIE xyY 1 19 White N 9.5/ 0.310 0.316 90.0 2 20 Neutral 8 N 8/ 0.310 0.316 59.1 3 21 Neutral 6.5 N 6.5/ 0.310 0.316 36.2 4 22 Neutral 5 N 5/ 0.310 0.316 19.8 5 23 Neutral 3.5 N 3.5/ 0.310 0.316 9.0 6 24 Black N 2/ 0.310 0.316 3.1 Table 20: Gretage MacBeth Basic Information Illuminances on the center of each tested surface were taken by IALD 61-680 which represent the overall illuminance for the targeted areas. Illuminance values were recorded. Canon Rebel Ti2 was used in the experiment. The camera settings were exactly the same as Test 1, as follow: White balance Daylight Aperture A-DEP (3.5, 4.5) File size 5184× 3456 Focus Auto Focal Length 18mm Sensor Size 22.3mm× 14.9mm Table 21: Test 2 Camera Settings 80 A series of three photos for each single light level setting were taken at the exposure levels: -2EV, 0EV, and +2EV. Figure 57: -2EV Exposure Photo of Illuminance at 4.9fc Figure 58: 0EV Exposure Photo of Illuminance at 4.9fc 81 Figure 59:+2EV Exposure Photo of Illuminance at 4.9fc The recorded luminance for each block under five different light levels as the following Table.22. Luminance Meter Readings Illuminance fc Block 1 cd/m2 Block 2 cd/m2 Block 3 cd/m2 Block 4 cd/m2 Block 5 cd/m2 Block 6 cd/m2 0.7 4.69 3.12 1.88 1.07 0.55 0.27 4.9 25.05 16.47 10.44 5.65 2.76 1.24 22.8 111.1 70.83 43.7 23.32 11.29 4.84 108 500.9 321 199.7 106.2 53.8 21.8 177.2 864.7 560.9 350.2 180.6 93.01 37.28 Table 22: Test 2 Luminance Meter Readings By processing the images in Photosphere, an HDR image for each illuminance level is generated. A luminance reading for each block is read in Photosphere. The reading is the average luminance in the center of each block which is about the same size as the view 82 field of the calibration Minolta Luminance Meter LS-110. The luminance readings from Canon T2i HDR images are recorded as follows: Illuminance fc Block 1 cd/m2 Block 2 cd/m2 Block 3 cd/m2 Block 4 cd/m2 Block 5 cd/m2 Block 6 cd/m2 0.7 4.66 2.79 1.7 0.977 0.393 0.181 4.9 26.3 17.1 9.87 5.64 2.4 1.02 22.8 117 73.7 41.3 22.2 9.11 3.45 108 595 365 209 110 47.4 16.2 177.2 842 547 317 174 80.5 27.7 Table 23: Canon T2i HDR Image Luminance Readings 83 Figure 60: Overall Comparison between Canon T2i HDR and Luminance Meter 84 Error is calculated in the way mentioned in Test 1: Illuminance fc Block 1 % Block 2 % Block 3 % Block 4 % Block 5 % Block 6 % 0.7 -0.64 -10.58 -9.57 -8.7 -28.55 -32.96 4.9 4.99 3.82 -5.46 -0.18 -13.04 -17.74 22.8 5.31 4.05 -5.49 -4.8 -19.31 -28.72 108 18.79 13.7 4.66 3.58 -11.9 -25.69 177.2 -2.63 -2.48 -9.48 -3.65 -13.45 -25.7 Table 24: Test 2 Canon T2i HDR luminance reading errors The error is plotted as following Figure 61 and Figure 62. As indicated by the results, HDR image luminance mapping performance for Block 3 and Block 4 is quite stable. The errors are within +10% or -10% range. Except for the case when light level on the center is about 108 fc, the errors are negative. For Block 1 and Block 2, except for one case: illuminance at 108 fc, the performance of HDR image is also consistent. The errors for Block 1 and Block 2 are within +10% or -10%, except the case described previously. However, the errors of Block 1 and Block 2 tend to be positive. For Block 5 and Block 6, the errors are all larger than -10%. Block 6 even got an error of -32.96% when the illuminance level is 0.7 fc. Block 6 has a larger error tendency than Block 5. But both Block 5 and Block 6 only have negative error which means the HDR luminance readings are always lower than luminance meter readings. Another noticeable fact is when the 85 light level is extremely low: 0.7 fc, the error for every block is negative and is comparatively larger than the other light levels. Thus, Canon T2i HDR image as a luminance mapping tool may not be very accurate under extremely low light levels. One possible explanation for this result could be the photonic sensitive capability of Canon T2i is not sufficient under low light level. Test 2 demonstrates that Canon T2i HDR image as a luminance has certain limitations: 1. Low reflectance surfaces intend to have larger errors toward the negative side, which means the luminance readings from HDR for low reflectance surfaces intend to be lower than luminance meter readings 2. When the absolute light level is extremely low, the HDR luminance readings intend to have negative errors, which means the luminance readings from HDR for low light level situations are intend to be lower than luminance meter readings. For most of the reflectance circumstances HDR luminance readings are quite accurate, the errors are between +10% and -10%. For glare analysis purposes, the HDR image is a possible and trustable luminance mapping tool. 86 Figure 61: Canon T2i HDR Luminance Error ( X axis is logarithm, log 2) 87 Figure 62: Canon T2i HDR Luminance Reading Error 3.3 Experiment Test 3: Software Each HDR generating software has its own algorithm in camera response curve producing, image aligning, and noise reduction. However, in order to eliminate the effects from image aligning and noise reduction, these two functions are blocked in all tests. Three kinds of software were applied, compared and analyzed in this paper: 1) Canon G12 Power Shot internalimplanted HDR function (software)& Adobe Photoshop, Canon G12 automatically takes three exposure bracketed images and merges them into one HDR image and then compresses and stores the HDR image in JPEG format; Adobe Photoshop 88 is used to read luminance values; 2) Bracket, it is a cross platform software which is considered to be less professional in luminance mapping; 3) Photosphere, was used in the previous Test 1 and Test 2 and it is the software used by most HDR analysis studies. 1. Canon G12 Power Shot internal HDR function + Adobe Photoshop Use the same scenario setting as used in Test 2, Canon G12 Power Shot was mounted on the tripod in the same position to take HDR photos of the color checker almost simultaneously as Canon T2i (switch these cameras). Figure 63: Canon G12 HDR Image Illuminance at 0.7 fc 89 The luminance meter readings are as following Table 25. Luminance Meter Illuminance fc Block 1 cd/m2 Block 2 cd/m2 Block 3 cd/m2 Block 4 cd/m2 Block 5 cd/m2 Block 6 cd/m2 0.7 4.69 3.12 1.88 1.07 0.55 0.27 4.9 25.05 16.47 10.44 5.65 2.76 1.24 22.8 111.1 70.83 43.7 23.32 11.29 4.84 108 500.9 321 199.7 106.2 53.8 21.8 177.2 864.7 560.9 350.2 180.6 93.01 37.28 Table 25: Luminance Meter Readings The process to get luminance value in Adobe Photoshop for HDR images is the same as described in Luminance Histogram Method (in Chapter 2). The HDR is imported into Adobe Photoshop. In Adobe Photoshop, the image is discarded of colors by selecting the Grayscale mode. Then the image is turned back to RGB mode. In histogram, luminosity channel a selected area’s mean luminosity could be read. 90 Illuminance fc Block 1 Block 2 Block 3 Block 4 Block 5 Block 6 0.7 202.11 178.89 153.56 120.96 79.12 43.48 4.9 233.89 211.4 178.86 135.32 91.55 46.31 22.8 233.17 211.72 178.28 132.93 90.51 44.51 108 236.63 213.3 181.69 134.48 91.44 36.99 177.2 238.39 213.92 182.71 134.33 88.61 33.88 Table 26: Canon G12 HDR Luminosity from Adobe Photoshop By linear calibration, the luminance could be calculated based on luminosity information recorded from Adobe Photoshop. Linear calibration is operated at a selected area. The calibration factor is the ratio between the luminance meter and the HDR luminosity value of that selected area. The whole image is then multiplied by the calibration factor. 1) Calibration at Block 1 (brightest area) Figure 64. Illuminance fc Block 1 cd/m2 Block 2 cd/m2 Block 3 cd/m2 Block 4 cd/m2 Block 5 cd/m2 Block 6 cd/m2 0.7 4.69 4.15 3.56 2.81 1.84 1.01 4.9 25.05 22.64 19.16 14.49 9.81 4.96 22.8 111.10 100.88 84.95 63.34 43.13 21.21 108 500.90 451.51 384.60 284.67 193.56 78.30 177.2 864.70 775.94 662.73 487.25 321.41 122.89 Table 27: Calibrated Luminance Reading of Canon G12 HDR Images 91 Error for each block under each light level is calculated as Test 2: Illuminance fc Block 1 % Block 2 % Block 3 % Block 4 % Block 5 % Block 6 % 0.7 0.00 33.05 89.54 162.33 233.82 273.69 4.9 0.00 37.47 83.49 156.51 255.26 299.99 22.8 0.00 42.42 94.38 171.60 281.98 338.18 108 0.00 40.66 92.59 168.05 259.78 259.18 177.2 0.00 38.34 89.24 169.79 245.57 229.64 Table 28: Canon G12 Photoshop Luminance Reading Error (Calibration at Block 1) The table is plotted as Figure 65. From the comparison of luminance between Canon G12 HDR image Adobe Photoshop readings and luminance meter readings in Figure 64the Canon G12 HDR image readings do not correspond to luminance meter very well. Generally, the luminance readings from Canon G12 HDR images are higher than luminance meter. 92 Figure 64: Overall Comparison between Canon G12 HDR and Luminance Meter Figure 65: Canon G12 HDR Luminance Reading Error (Calibration at Block 1) 0.00% 50.00% 100.00% 150.00% 200.00% 250.00% 300.00% 350.00% 400.00% 0 50 100 150 200 Error Illuminance on the center fc Canon G12 HDR error Block 1 Block 2 Block 3 Block 4 Block 5 Block 6 93 By looking at the error plot, it is easy to see that all the errors are positive (except the Block 1 ones, they are zero because these areas are the calibration areas). The error range is between 33.05% and 338.18% which is very high. For Block 5 and Block 6, the errors have a tendency of decreasing as the illuminance level is increasing, but the Block 2, Block 3 and Block 4 errors are quite consistent over the different light levels. The error is increasing when the reflectance is decreasing from Block 2 to Block 6 over a single light level, in other words within one HDR image. This could be caused by the camera automatically compresses the HDR image into JPEG format which has only 8 bit stored information about luminosity. 8 bit only has the range from 0-255. 2) Calibration at the Block 6 (Darkest area) Figure 66. Redline is the ideal trendline: Illuminance fc Block 1 cd/m2 Block 2 cd/m2 Block 3 cd/m2 Block 4 cd/m2 Block 5 cd/m2 Block 6 cd/m2 0.7 1.26 1.11 0.95 0.75 0.49 0.27 4.9 6.26 5.66 4.79 3.62 2.45 1.24 22.8 25.35 23.02 19.39 14.45 9.84 4.84 108 139.46 125.71 107.08 79.26 53.89 21.80 177.2 262.31 235.39 201.05 147.81 97.50 37.28 Table 29: Calibrated Luminance Reading of Canon G12 HDR Images 94 The calculated error for the Canon G12 HDR luminance errors (Figure 67): Illuminance fc Block 1 % Block 2 % Block 3 % Block 4 % Block 5 % Block 6 % 0.7 -73.24 -64.40 -49.28 -29.80 -10.67 0.00 4.9 -75.00 -65.63 -54.13 -35.87 -11.18 0.00 22.8 -77.18 -67.50 -55.64 -38.02 -12.83 0.00 108 -72.16 -60.84 -46.38 -25.37 0.17 0.00 177.2 -69.66 -58.03 -42.59 -18.16 4.83 0.00 Table 30: Canon G12 Photoshop Luminance Reading Error (at Block 6) The absolute difference between Canon G12 HDR luminance readings and luminance meter readings is very large. When the standard luminance increases, the difference increases dramatically. Especially when at the Block 1 and under 177.2 fc light level, the Canon G12 HDR image luminance reading is 262.31 cd/m2 while the standard luminance meter reading is 864.7 cd/m2. When considering about the errors, the error range is smaller than Calibration at Block 1. The range is from -77.18% to 0.17%. This is because the comparatively larger differences happen at comparatively higher absolute standard luminance side. It is quite obvious that generally, the errors are negative and increasing when the reflectance (luminance) is increased across a single image. This is just opposite to the result of Calibration at Block 1. The limitation of 8 bit luminosity information storage may be the reason. It is also possible that the image is being intentionally “improved” by the software to reduce the range. 95 Figure 66: Overall Comparison between Canon G12 HDR and Luminance Meter Figure 67: Canon G12 HDR Luminance Reading Error (Calibration at Block 6) 96 2. Bracket In this case, the same photos taken by Canon T2i in Test 2 were processed in Bracket. Select Exposure Variation by Read from EXIF and Camera Response by Recover New Response for Canon T2i. Then read the luminance value by pixel rather than read mean luminance for a selected area. The tested surfaces (color checker blocks) are quite diffused and the light levels are quite even. The pixel luminance values are similar over the luminance meter viewport area. So the center pixel of each pixel was selected to represent the luminance meter viewport area. The luminance values are read in Bracket and recorded as following: Illuminance fc Block 1 cd/m2 Block 2 cd/m2 Block 3 cd/m2 Block 4 cd/m2 Block 5 cd/m2 Block 6 cd/m2 0.7 0.14 0.0806 0.0527 0.0302 0.018 0.0084 4.9 0.661 0.441 0.274 0.147 0.081 0.0421 22.8 2.755 1.723 1.225 0.545 0.315 0.144 108 14.672 8.408 5.9 2.802 1.638 0.702 177.2 21.984 14.024 9.176 4.638 2.816 1.114 Table 31: Canon T2i Bracket HDR Luminance Readings The luminance values are quite different from luminance meter readings. So linear calibration is needed. 97 1) Calibration at Block 1, Figure 85: Illuminance fc Block 1 cd/m2 Block 2 cd/m2 Block 3 cd/m2 Block 4 cd/m2 Block 5 cd/m2 Block 6 cd/m2 0.7 4.69 2.7001 1.76545 1.0117 0.603 0.2814 4.9 25.05 16.71263 10.38381 5.570877 3.069667 1.595469 22.8 111.1 69.48287 49.40018 21.97804 12.7029 5.807042 108 500.9 287.0479 201.4252 95.65988 55.92109 23.96618 177.2 864.7 551.6081 360.921 182.4272 110.7622 43.81713 Table 32: Calibrated Canon T2i Bracket HDR Luminance Readings (at Block 1) The errors for calibrated Canon T2i Bracket HDR images are calculated as following Illuminance fc Block 1 % Block 2 % Block 3 % Block 4 % Block 5 % Block 6 % 0.7 0.00 -13.46 -6.09 -5.45 9.64 4.22 4.9 0.00 1.47 -0.54 -1.40 11.22 28.67 22.8 0.00 -1.90 13.04 -5.75 12.51 19.98 108 0.00 -10.58 0.86 -9.92 3.94 9.94 177.2 0.00 -1.66 3.06 1.01 19.09 17.54 Table 33: Calibrated Canon T2i Bracket HDR Luminance Readings Errors (at Block 1) 98 Figure 68: Overall Comparison of Calibrated Bracket HDR and Luminance Meter 99 Figure 69: Canon T2i Bracket HDR Luminance Reading Error From the overall comparison diagram, the calibrated Canon T2i Bracket HDR luminance readings over luminance meter readings, the tendency is very close to the ideal trendline. The overall luminance errors are in the range between -13.46% and 28.67%. The majority of errors for Block 2, Block 3, Block 4 and Block 5 are within -15% to +15% range. For Block 6, most of the errors are more than 10% and always in the positive side. This result indicates that Canon T2i Bracket HDR image as a luminance mapping tool is quite reliable for most of the reflectance circumstances. However, it is very interesting that almost of Block 2 errors are in the negative side and the lower the reflectance the surface is the larger the error is (absolute value). Block 3 and Block 4 have better correlation with luminance meter than the rest blocks. 100 2) Calibration at Block 6, Figure 70 (redline is the ideal trendline): Illuminance fc Block 1 cd/m2 Block 2 cd/m2 Block 3 cd/m2 Block 4 cd/m2 Block 5 cd/m2 Block 6 cd/m2 0.7 4.50 2.59 1.69 0.97 0.58 0.27 4.9 19.47 12.99 8.07 4.33 2.39 1.24 22.8 92.60 57.91 41.17 18.32 10.59 4.84 108 455.63 261.10 183.22 87.01 50.87 21.80 177.2 735.69 469.31 307.07 155.21 94.24 37.28 Table 34: Calibrated Canon T2i Bracket HDR Luminance Readings The errors for calibrated Canon T2i Bracket HDR images are calculated as following: Illuminance fc Block 1 % Block 2 % Block 3 % Block 4 % Block 5 % Block 6 % 0.7 -4.05 -16.96 -9.90 -9.28 5.19 0.00 4.9 -22.28 -21.13 -22.70 -23.37 -13.56 0.00 22.8 -16.65 -18.24 -5.78 -21.45 -6.22 0.00 108 -9.04 -18.66 -8.25 -18.07 -5.45 0.00 177.2 -14.92 -16.33 -12.31 -14.06 1.32 0.00 Table 35: Calibrated Canon T2i Bracket HDR Luminance Readings Errors From the overall comparison between calibrated Canon T2i Bracket HDR luminance (at Block 6) and luminance meter diagram, it is noticeable that when the standard luminance increases, the differences between HDR luminance readings and luminance meter readings are increasing. But because of the comparatively larger absolute luminance values, the errors are comparatively smaller. From Table 35, it is easy to tell that the error 101 range for HDR image luminance readings is from -23.37% to 5.19%. The majority of HDR luminance errors are negative which is quite opposite to the calibration at Block 1. It is quite interesting that Block 2 and Block 4 have larger errors than Block 1, which would be predicted to have the most magnificent errors. Figure 70: Overall Comparison of Calibrated Bracket HDR and Luminance Meter 102 Figure 71: Canon T2i Bracket HDR Luminance Reading Error 3. Photosphere Photosphere is the software used in Test 2. So the result of Test 2 is the result here: Illuminance fc Block 1 cd/m2 Block 2 cd/m2 Block 3 cd/m2 Block 4 cd/m2 Block 5 cd/m2 Block 6 cd/m2 0.7 4.66 2.79 1.7 0.977 0.393 0.181 4.9 26.3 17.1 9.87 5.64 2.4 1.02 22.8 117 73.7 41.3 22.2 9.11 3.45 108 595 365 209 110 47.4 16.2 177.2 842 547 317 174 80.5 27.7 Table 36: Canon T2i Photosphere HDR Luminance Readings 103 Calibrate the image at Block 1: Illuminance fc Block 1 cd/m2 Block 2 cd/m2 Block 3 cd/m2 Block 4 cd/m2 Block 5 cd/m2 Block 6 cd/m2 0.7 4.69 2.81 1.71 0.98 0.40 0.18 4.9 25.05 16.29 9.40 5.37 2.29 0.97 22.8 111.10 69.98 39.22 21.08 8.65 3.28 108 500.90 307.27 175.95 92.60 39.90 13.64 177.2 864.70 561.75 325.55 178.69 82.67 28.45 Table 37: Calibrated Canon T2i Photosphere HDR Luminance Readings The calculated errors for calibrated Photosphere HDR luminance readings are as Table 38: Illuminance fc Block 1 % Block 2 % Block 3 % Block 4 % Block 5 % Block 6 % 0.7 0.00 -10.00 -8.99 -8.10 -28.09 -32.53 4.9 0.00 -1.11 -9.95 -4.92 -17.18 -21.65 22.8 0.00 -1.20 -10.26 -9.60 -23.38 -32.31 108 0.00 -4.28 -11.89 -12.80 -25.83 -37.44 177.2 0.00 0.15 -7.04 -1.06 -11.12 -23.69 Table 38: Calibrated Canon T2i Photosphere HDR Luminance Readings Errors 104 Illuminance fc Block 1 % Block 2 % Block 3 % Block 4 % Block 5 % Block 6 % 0.7 -0.64 -10.58 -9.57 -8.7 -28.55 -32.96 4.9 4.99 3.82 -5.46 -0.18 -13.04 -17.74 22.8 5.31 4.05 -5.49 -4.8 -19.31 -28.72 108 18.79 13.7 4.66 3.58 -11.9 -25.69 177.2 -2.63 -2.48 -9.48 -3.65 -13.45 -25.7 Table 39: Non-calibrated Canon T2i Photosphere HDR Luminance Readings Errors By comparing the calibrated HDR image luminance errors with the non-calibrated errors (Table. 39), it is easy to find that calibration does not necessarily minimize errors. When the errors within a single HDR images are all negative or positive, by calibrating the Block 1, the error range and error for each block can be decreased and the decrease for each block depends on the error of Block 1. For example, when the light levels are 0.7 fc and 177.2 fc, the errors of Photosphere HDR images for all six blocks are negative and Block 1 errors under 0.7 fc light level and 177.2 fc light level are -0.64% and -2.63% respectively. After the calibration at Block 1, the errors decrease for 0.7 fc light level is much smaller than 177.2 fc light level and almost proportion to the before calibration Block 1 error. When the errors in a single image are scattered above and below zero, the calibration at Block 1 may increase the error for some blocks rather than decrease. This is because under one set of light level, the Photosphere luminance readings may be larger than the luminance meter readings in some blocks (usually the high reflectance ones) but 105 smaller in other blocks (the lower reflectance one). If calibrate the Block 1 (in other words, decrease the error of Block 1 to 0), negative errors (HDR luminance smaller than luminance meter readings) of the lower reflectance blocks will be exaggerated. The main reason for this result is the camera response curve generated from Photosphere is slightly different from the luminance meter -- Minolta Luminance Meter LS-110’s photonic response curve and the difference non-linear. The comparison among Canon G12 Power Shot inplant HDR function + Adobe Photoshop (calibrated at Block 1), Canon T2i Bracket HDR image (calibrated at Block 1) and Canon T2i Photosphere HDR image can be made in the following diagram Figure 72: Comparison of Three HDR Image Software. Canon G12 Power Shot internal HDR function + Adobe Photoshop has the worst performance as a luminance mapping tool. Both Bracket and Photosphere work well. Surprisingly, calibrated Bracket HDR image has better performance as illustrated in Figure 72, or by comparing the error range (error range for HDR image which is calibrated at the Block 1 (white) is -13.46%~28.67%; error range for HDR image which is calibratet Block 6 (black) is -22.37% ~5.19% ) than Photosphere (error range for non-calibrated HDR image is -32.96% ~ 18.79%). However, Photosphere needs no calibration and works well as a luminance mapping tool. 106 Figure 72: Comparison of Three HDR Image Software 0 100 200 300 400 500 600 700 800 900 1000 0 100 200 300 400 500 600 700 800 900 1000 HDR Imgae Lumiance Reading cd/m2 Standard Luminance cd/m2 (Luminance Meter) Comparison of Three HDR Software Luminance Readings and Luminance Meter Readings Canon G12 HDR Function + Adobe Photoshop (Calibrated at Block 6) Canon T2i Bracket HDR (Calibrated at Block 1) Canon T2i Photosphere HDR (without Calibration) 107 Chapter 4: Three Glare Analysis by Using HDR Images 4.1 Experiment Model Tests 4.1.1 Model Settings Since the Glare Indices are all developed for artificial light sources and an interior environment, the glare sources are relatively small in the overall viewed image (point sources). So the application of Glare Index: Unified Glare Rating (UGR) and BRS Glare Equation (BRS/ BGI) to wider glare source may not be very accurate. The model is made of foam board with black exterior and white interior. The dimension for this model is: 20in (width)× 30in (length)× 20in (height). The light sources in these two tests are 9W compact fluorescent bulb and 150W tinted incandescent lamp, outside of the box. The light comes through the holes cut in the wall, which are covered by translucent paper. A skylight is opened on the top of the box and it is covered by same translucent paper. However, there is no direct light source from the skylight. It is merely the secondary light source for the tested model (interior)..The luminance within the box is controlled by moving the lamp towards or away from the box. Figure 73 and Figure 74. 108 Figure 73: The Front Side (with Windows) of the Model Figure 74: The Back Side of the Model and Camera is Mounted on a Tripod 109 Use Canon T2i to take photos. Settings are as following: White balance Daylight Aperture Manual f/14 File size 3456× 2304 (M) Focus Auto Focal Length 18mm Sensor Size 22.3mm× 14.9mm Table 40: Canon T2i Settings Use Photosphere to generate HDR images. Two distinct light levels were made by applying 9W compact fluorescent lamp and 150W tinted incandescent lamp at the same position. Test 1: 9W Compact Fluorescent lamp Since the direct light source is within close distance, the absolute luminance values at the holes are very high. Meanwhile, since the light penetrates through the holes are the preliminary light source in the tested model room. The room is quite dim. So a series of 9 stops of photos were taken to fully cover dynamic range of luminance across the scene as follows: 110 Figure 75: Equivalent to -6EV Figure 76: Equivalent to -5EV Figure 77: Equivalent to -4EV Figure 78: Equivalent to -3EV Figure 79: Equivalent to -2EV Figure 80: Equivalent to -1EV 111 Use Photosphere to generate HDR image, save it as RGBE file (.hdr). Name the file as Test10.hdr Test 2: 150W Tinted Incandescent Lamp Since the absolute luminance in the holes are even higher than Test 1, the photos taken for making HDR image are in a larger range of luminance stops. A totally 14 stops of Figure 81: Equivalent to 0EV Figure 82: Equivalent to +1EV Figure 83: Equivalent to +2EV 112 photos were taken and use every other stop to make an HDR image, which is called Test11.hdr. 113 Figure 84: Equivalent to -10EV Figure 85: Equivalent to -8EV Figure 86: Equivalent to -6EV Figure 87: Equivalent to -4EV Figure 88: Equivalent to -2EV Figure 89: Equivalent to 0EV 114 Figure 90: Equivalent to +2EV 4.1.2 GlareIndices (C++ Program) Glare Analysis Use GlareIndices (C++ Program). Input the camera parameters and follow the steps in Chapter 2. 115 Results: Test 1: 9W Compact Fluorescent lamp Figure 91: Test 1 Histogram in Excel BGI=14.29778 Just Perceptible UGR=13.08286 Just Perceptible Schiler Index= N/A No Glare No Spike in Histogram, so no glare in Schiler Method. 116 Test 2: 150W Tinted Incandescent Lamp Figure 92: Test 2 Hitogram in Excel BGI=23.56508 Uncomfortable UGR=23.43893 Just Uncomfortable Schiler Index= 204 High Possibility of Glare By following the steps in Chapter 2, open the generated test11.csv with Microsoft Excel. Select the first two columns and make a histogram. Since the number overall pixels in the image is 7,962,624. It is recommended to use log2 scale at the Y axis. The result is show in Figure 92. Then define the Bell shape minimum luminance: 0 cd/m2, maximum 311 cd/m2 and the Spike shape minimum luminance: 373 cd/m2 and maximum 7340 cd/m2. Input these information to C++ (GlareIndices) and start Process2. Schiler Index is 204. 117 According to Luminance Histogram Method, there is high possibility of glare in the scene. 4.1.3 Radiance Glare Analysis Process the same HDR images in Radiance. Use command “findglare” to generate .glr file. Then use command “glarenx” to analysis .glr file and calculate the glare indices. Since Radiance “findglare” is defaulted to calculate a fish-eye lens view, a normal lens is used, Radiance will give out a warning and calculate the missing sample area. Radiance will fill the missing sample with mean luminance value of the image. 118 The results: Test 1: Test10.hdr 9W Compact Fluorescent Lamp Figure 93: Test 1 Radiance Screenprint Missing 74% of samples BGI/BRS= 12.409648 Imperceptible UGR=11.167304 Just Perceptible 119 Test 2: Test11.hdr 150W Tinted Incandescent Lamp Figure 94: Test 2 Radiance Screenprint Missing 74% of samples BGI/BRS= 23.307489 Uncomfortable UGR=23.295094 Just Uncomfortable 4.1.4 Comparison between GlareIndices (C++ Program) and Radiance Results The GlareIndices (C++ Program) and Radiance results (values) are slightly different but quite close. Generally, GlareIndices (C++ Program) results (values) are larger than the Radiance results. This may be caused by three possible reasons: 1. The luminance calculation formulae (based on Radiance RGBE HDR file) are slightly different in 120 GlareIndices and Radiance; 2. Radiance has the function of “Merge the Source” which may cause some glare pixels (or areas smaller than 0.005 steradian disk area on the image) absorbed by the background; 3. The glare formulae for calculations are slightly different (eg. The position index P is different from Radiance). However, these differences are not crucial. When applying glare criterion, Test 2 gives exactly same glare analysis results. However, in Test 1, GlareIndices (C++ Program) is predicting Just Perceptible in both BGI and UGR but Radiance is slightly different with Imperceptible for BGI and Just Perceptible for UGR. 4.1.5 Comparison among Three Glare Indices: BGI, UGR and Schiler Index For both Test 1 and Test 2, BGI index and UGR index results agree with each other. This may due to the two indices are derived from the same formula: But Luminance Histogram Method gives quite different glare prediction than BGI or UGR. In Test 1, by looking at the histogram, since there is no obvious spike, so the conclusion of no glare is drawn. But it is quite different from BGI and UGR. And in Test 2, it is a little bit tricky to define the spike shape. With a largely expanded luminance range (than Adobe Photoshop’s 0-255), and enormous pixels (3456× 2304) counted in the 121 image, the spike is not clearly shaped in the histogram. And with author’s definition, a conclusion of high possibility of glare is generated. The person who is using the Luminance Histogram Method is influencing the glare prediction. It is good because glare is people’s sensation, so it seems to be good when human’s decision is involved. However, it may be not very convincing since different people will generate quite different glare prediction for a single scene and this glare prediction may not represent the majority of people’s opinion. 4.2 Interior Test 4.2.1 Interior Settings The primary light source is the florescent mounted on the ceiling. White frosted light fixture diffused the light. Daylight is also penetrating through the blades. It is diffused daylight because the window is facing the north. 122 The camera used is Canon Rebel T2i, and the settings are: White balance Daylight Aperture A-DEP (3.5) File size 3456× 2304 (M) Focus Auto Focal Length 18mm Sensor Size 22.3mm× 14.9mm Table 41: Canon T2i Settings Since the light source is directly observed (shot by the camera), the luminance of the light is very high. So it is very important to make sure the luminance is accurately captured which means there is no saturate pixels in light source. So for this situation, the shortest exposure photo was taken at -7EV. A serious of 8 stops of photos were taken. Use Photosphere to process the 8 photos and generate a Radiance RGBE file called test7.hdr. 123 Figure 95: Equivalent to -7EV Figure 96: Equivalent to -5EV Figure 97: Equivalent to -3EV Figure 98: Equivalent to -1EV Figure 99: Equivalent to +1EV Figure 100: Equivalent to +3EV 124 Figure 101: Equivalent to +5EV Figure 102: Equivalent to +7EV 4.2.2 GlareIndices(C++ Program) Glare Analysis By using the GlareIndices (C++ Program), the luminance histogram could be generated in Excel, as in Figure 103. The glare indices BGI and UGR are calculated. Figure 103: Interior Test Histogram in Excel (Y is scaled at log2) BGI=19.68104 Just Acceptable UGR=19.98207 Just Acceptable 125 It is very tricky to use Luminance Histogram Method under this situation. If no spike is defined, this means no glare in the scene. If a spike is defined as the Figure 103 shows: BellMax at 175 and BellMin at 0, SpikeMax at 3897 and SpikeMin at 1250, then the Schiler Index is 300, that means there is extreme glare in the scene. 4.2.3 Radiance Glare Analysis Analysis the same test7.hdr in Radiance, BGI /BRS and UGR are calculated as in Figure 104 shows: Figure 104: Interior Test Radiance Screenprint BRS/BGI=18.878785 Perceptible UGR=19.212521 Just Acceptable 126 4.2.4 Comparison between GlareIndices (C++ Program) and Radiance Results In Interior Test, GlareIndices (C++ Program) and Radiance Results are very close. The GlareIndices (C++ Program) also gives higher glare ratings than Radiance. When applied with glare criterion, BGI in GlareIndices (C++ Program) is giving more serious glare prediction. This is because, the criterion between Just Acceptable and Perceptible is 19 which is close to both GlareIndices (C++ Program) and Radiance and just in between. Also in GlareIndices (C++ Program) gives consistent glare prediction from BGI and UGR, but Radiance is different for the same scene. 4.2.5 Comparison among Three Glare Indices: BGI, UGR and Schiler Index BGI and UGR give quite coincided glare predictions. But Luminance Histogram Method gives much severe glare prediction. The luminance histogram of Interior Test has clear spike shape and bell shape. The definition of the spike shape and bell shape for the scene is not a crucial reason for quite distinct glare prediction. The reason for less severe glare prediction for a high contrast scene may be the position of the glare source. The glare source in this scene is the lamp on the ceiling. It is not in the focus of the scene. In Luminance Histogram Method, the position of light source (possible glare source) is not considered. So the high contrast of luminance in the scene leads to an extremely high Schiler Index which indicates a very serious glare issue. 127 4.3 Exterior Night Test 1 4.3.1 Test Settings The light sources in the exterior night test scene are compact fluorescent lamps. The camera used in the test is Canon Rebel T2i as before. The camera settings are: White balance Daylight Aperture A-DEP (3.5) File size 3456× 2304 (M) Focus Auto Focal Length 18mm Sensor Size 22.3mm× 14.9mm Table 42: Canon T2i settings Since it is a night shot with direct light source in scene, the luminance range in the scene is extremely wide. When the camera was set to A-DEP mode to take a series of bracketed exposure images as follows: 128 Figure 105: -2EV Figure 106: -1EV Figure 107: 0EV Figure 108: +1EV Figure 109: +2EV The shortest exposure photo still got saturated pixels in the lamp areas which are the light sources. Because of the over-exposure at the light source areas, which on the other hand 129 are the possible glare sources, the light sources luminance captured by the photos is not accurate. The luminance recorded in the photos is lower than what they actually are. If we process these photos in Photosphere and generate the HDR image called sce3.hdr, and then apply GlareIndices (C++ Program) and Radiance to analyze it, the results are not reliable. 4.3.2 GlareIndices (C++ Program) Glare Analysis By processing sce3.hdr in GlareIndices (C++ Program), the results are as following Figure 110 shows: Figure 110: Exterior Night Test 1 Histogram in Excel (Y axis is scaled at log2) BGI= 3.28553 N/A no glare UGR=-0.85796 N/A no glare 130 By defining the BellMax at 21, BellMin at 0, SpikeMax at 32 and SpikeMin at 30 in Luminance Histogram Method, a Schiler Index of 68 is calculated. 4.3.3 Radiance Glare Analysis Process the same sce3.hdr in Radiance as following Figure 111 shows: Figure 111: Exterior Night Test 1 Radiance Screenprint BGI/BRS=0.732661 N/A no glare UGR=-2.269183 N/A no glare 131 4.3.4 Comparison GlareIndices (C++ Program) and Radiance Results GlareIndice (C++ Program) and Radiance results are slightly different. As mentioned before, the results are not reliable since the possible glare sources’ luminance values are smaller than their actual luminance. And the extent of decreasing of luminance is not perceivable. By that means, the glare indices BGI and UGR are accordingly smaller than the actual values. The negative results are mathematically theoretically possible but not reasonable in the scene. 4.3.5 Comparison among Three Glare Indices: BGI, UGR and Schiler Index BGI and UGR results are not able to give accurate glare prediction since the decreased light source luminance. And the BGI and UGR are very small and should be smaller than what they should be. And the relatively smaller Schiler Index also reflects the same tendency. However, Luminance Histogram Method gives quite opposite glare prediction. Schiler Index is 68 which is predicting a very high possibility of glare in the scene. 132 4.4 Exterior Test 2 4.4.1 Test Settings The test settings are exactly the same as Exterior Test 1. Canon Rebel T2i is set to be manually controlled. A series of 16 exposure stops photos were taken. Apply every other stop photo into Photosphere and generate the HDR image called ExteriorTest.hdr. 133 Figure 112: Equivalent to -11EV Figure 113: Equivalent to -9EV Figure 114: Equivalent to -7EV Figure 115: Equivalent to -5EV Figure 116: Equivalent to -3EV Figure 117: Equivalent to -1EV 134 Figure 118: Equivalent to +1EV Figure 119: Equivalent to +3EV 4.4.2 GlareIndices (C++ Program) Glare Analysis Figure 120: Exterior Night Test 2 Histogram in Excel (both in logarithm) 135 Figure 121:Exterior Night Test 2 Histogram in Excel (Linear) BGI=17.4109 Perceptible UGR=14.2315 Just Perceptible Open ExteriorTest.csv file with Microsoft Excel. If the histogram is plotted in a linear scale as Figure 121 shows. The histogram seems to be an “L” shape which is squeezed into the left side. It means that the majority of pixels in the HDR image are towards the low luminance. There is no obvious bell or spike shape in the histogram. However, if the histogram is plotted in a logarithm scale as Figure 120 shows the shape is more readable. Studying on the histogram it is easy to define the bell shape but the spike is indistinctive. As in the study, the author defines the Bell shape maximum luminance 88 cd/m2 and minimum 0 cd/m2; Spike shape maximum luminance 1248 cd/m2 and minimum 96 cd/m2. Schiler Index is 9281 which indicate there is high possibility of glare in the scene. 1 1000001 2000001 3000001 4000001 5000001 6000001 7000001 8000001 0 5000 10000 15000 20000 25000 30000 136 4.4.3 Radiance Glare Analysis Figure 122: Exterior Night Test 2 Radiance Screenprint BGI/BRS=8.941122 (N/A) No Glare UGR=6.461681 (N/A) No Glare 4.4.4 Comparison GlareIndices (C++ Program) and Radiance Results GlareIndices (C++ Program) gives much higher glare rating than Radiance. The reason is highly possible be the No. 2 reason described in Chapter 4.1.4: 2. Radiance has the function of “Merge the Source” which may cause some glare pixels (or areas smaller than 137 0.005 steradian disk area on the image) absorbed by the background. Since the light source is very small in the scene and neglecting glare pixels is going to infect the glare calculation in a larger extend for a bright small light source than a relatively large light source, such as the Interior Test. 4.4.5 Comparison among Three Glare Indices: BGI, UGR and Schiler Index In overall, BGI and UGR in both GlareIndices (C++ Program) and Radiance are coincided (within the same software). Luminance Histogram Method gives even greater glare rating. The Schiler Index is remarkable 931. Since the Luminance Histogram Method glare criterion (ratio) is 3, which is comparatively smaller than the Schiler Index for this scene. 4.5 Analysis on Glare Analysis by Using HDR Images 4.5.1 Comparison between GlareIndices (C++ Program) and Radiance The GlareIndices (C++ Program) and Radiance results (values) are slightly different but quite close for most of the circumstances. When applied to glare criterion, the results are coincided for most of the scenarios. Since Radiance is professional and trustworthy software used by many researchers and plug-in to other design, rendering and analysis software. The author believes Radiance gives reliable glare indices results based on any 138 given HDR images. As mentioned earlier in Chapter 2, Radiance analyzes images captured or rendered in fisheye lens. When Radiance is calculating glare indices for non-fish-eye lens, it automatically fills the missing samples with the average luminance over the scene. This is not crucial to the glare indices since it is not going to change the background luminance nor the light (glare) source luminance. Radiance results are relatively smaller than GlareIndices (C++ Program). This may be caused by 1. Slightly different algorism of luminance value calculation; 2. Radiance defaulted to neglect small glare sources to background; 3. The glare formulae for calculations are slightly different. From the tested scenarios, GlareIndices (C++ Program) performs very well for HDR image glare analysis. So as a PC glare analysis program which is designed for non-fish-eye lens, GlareIndices (C++ Program) is reliable. But if is better to neglect small light sources or isolated high luminance pixels or keep every pixel’s luminance as they were recorded and calculate the glare indices based on every possible glare source (high luminance pixels) needed to be further studied. 4.5.2 Analysis of HDR Images as a Glare Analysis Tool As Test 3 demonstrates that an accurate HDR image is very important for the following glare analysis. Although in Chapter 3, a series of tests have been done to verify that HDR image as a luminance mapping tool is reliable, an HDR image is not necessarily accurate. 139 Only correctly generating an HDR image for any studied scene could be used for glare analysis. One needs to make sure the HDR image is in good quality without too much noise, image is correctly aligned, most importantly no information missing. The information missing means not all the luminance information in the scene are recorded in the HDR image. It usually happens at the light source it gets over exposed. Make sure in the shortest exposed photo, no blasted pixels, which means, no pixel is saturated. In cameras, there is histogram display mode. One can check the histogram presented for the shortest exposed photo for no pixel at 255 or the twinkling warning on the image. However, since the display LCD screens are usually too small to check. It is highly recommended that take 2 stops more when the histogram for the shortest exposed photo seems to be clear of saturated pixels. Then before process the bracketed photos in Photosphere. Check the shortest exposed image in Photoshop. Box selecting the light sources which could be the possible glare sources, make sure there is no pixel counted for luminosity of 255. Only by that means, the light source luminance could be recorded accurately. And for the over exposed end, in the longest exposed image, each pixel’s luminance should be above 25 to get rid of possible noise defection in images. 140 4.5.3 Glare Indices UGR and Luminance Histogram Method By comparing the Glare Indices UGR and BRS in the tests, it is easy to find out that they are corresponding with each other very well as they are expected to be. However, Luminance Histogram Method does not correspond with the other two glare indices very well. There are possibly two major reasons: 1. Luminance Histogram Method does not consider the position of light/ possible glare source; 2 Luminance Histogram Method is developed based on low dynamic luminosity range images rather than real world luminance range. For example in the Interior Test, the lamp is very bright and has an average luminance about 3500 cd/m2 in the center and the average of background luminance is about 12 cd/m2. But since the lamp is not in the focus of the scene/ viewer, the glare may not be that obvious or serious. Also, because of the extended luminance range, sometimes it is hard to define a spike. With the largely expanded luminance range, for example in Test 2, the luminance range is 3897/0.053 rather than 255/1. The Schiler Index is enlarged because of the expanded luminance range. So the criterion of possible glare 3:1 may not plausible anymore. Absolute luminance range may affect the Schiler Index criterion. Proposed Luminance Histogram Method criterion is connected with the absolute luminance range in the scene and the criterion may be a floating number or a formula. For example, in an HDR with a 141 luminance range of 0~4000 cd/m2, it should be 250, in a range of 0~30 cd/m2, it should be 100. However, the specific Luminance Histogram Method criterion need to be further studied with a psychological statistic survey. On the other hand, Luminance Histogram Method’s accuracy may be limited because of its neglecting of light sources positions. It may be limited in application for glare prediction of a task area or any small and close scenario where position index is not significantly influence the glare prediction. 142 Chapter 5: Conclusion 5.1 HDR Image as a Luminance Mapping Tool The use of HDR image based luminance mapping tool to measure luminance within the field of view provides a remarkable potential for improved understanding between measurements and users response. Compared with the traditional luminance meter recording tool, HDR image has the benefits of time efficiency and less human operation mistakes involved. In the series of experiments carried out in Chapter 3, 1. Camera: As demonstrated in the tests, the digital camera used for image capture is a key influencing factor. The tests demonstrated that a larger photonic sensor size digital camera is prone to give more accurate luminance readings. Compact digital cameras are prone to give smaller luminance readings since their photonic sensor are much smaller than the luminance meter. 2. Luminance reading surfaces: Low reflectance surfaces incline to generate larger errors in luminance mapping. The low reflectance surfaces’ HDR image luminance readings are prone to be smaller than the luminance meter readings for Canon Rebel T2i. This may be because of the different relative spectral 143 responsive of Konica Minolta luminance meter and Canon Rebel T2i. So, it may be camera independent. But it also could be other reasons which may limit the accuracy of HDR image luminance mapping. Since the low reflectance surfaces are intended to be the darker areas within a scene which are possibly the background in glare analysis (which are less likely to be the glare sources), lower luminance readings may increase the severity of glare prediction. 3. Software: Different HDR image creating software may have different accuracy performance. Canon G12 Power Shot internal HDR function is proved to be not useful for luminance mapping purposes. The function is primarily designed for artistic photography performance rather than scientific luminance recording since it compress the HDR image to a JPGE file (which can only store as much as luminosity information as the image used for creating the HDR image). Photosphere is the HDR image creating software used by most of the professional papers. Its performance as a luminance mapping tool is acceptable. The error range is within the expectation. Bracket is the software with less scientific professional attention than artistic photographic application. However, in the experiment, its performance as a HDR making software for luminance purposes is surprisingly though slightly superior to Photosphere when it is calibrated. But its 144 disadvantage is it needs to be calibrated which means it cannot work by itself as a luminance mapping tool to substitute luminance meter. 4. Calibration: Although calibration is mentioned by almost every HDR image luminance mapping manual, it is demonstrated by the experiments that it may not be necessary or proper for some software but be the essential and crucial factors for other software. For the tested Canon G12 Power Shot internal HDR, it is necessary to calibrate the luminosity values in Photoshop with real world luminance meter readings. For Bracket, the absolute luminance values directly read from the HDR image are not accurate, but calibration can improve the luminance mapping performance. So calibration is proper and necessary for more accurate luminance readings. And it is found that to calibrate at bright (high reflectance) surface is better than to calibrate at dark (low reflectance) surface. For Photosphere, it is found that calibration is not necessary or even not recommended. The calibration may decrease the error for certain portions of the image but increase the error in other portion. And without calibration, the error range of Photosphere HDR images is acceptable and the performance as a luminance mapping tool is trustable. No calibration is one advantage of Photosphere HDR image since the primary goal of HDR image as a luminance 145 mapping tool is diminishing the use of luminance meter in luminance recording for a scene. 5.2 Glare Analysis Based on HDR images 5.2.1 GlareIndices (C++ Program) GlareIndices (C++ Program) is created for exclusively glare analysis based on Radiance RGBE formatted HDR image. It is a user-friendly software. It is run on the Windows system. The glare analysis results are three popular glare analysis indices ratings: BGI, UGR and Schiler Index (for Luminance Histogram Method). By comparing the results of a series of diverse test cases with Radiance which is the dominant glare analysis software, the performance of GlareIndices (C++ Program) is quite reliable. Only in the Exterior Test, there is sharp difference between GlareIndices (C++ Program) and Radiance. As mentioned in Chapter 4, the reason may be the neglecting of small glare pixels algorithm in Radiance. However, if it is more accurate to neglect the small glare pixels in glare analysis by using HDR images need to be further studied by more tests with psychological surveys. 146 5.2.2 GlareIndice (C++ Program) BGI and UGR In GlareIndice (C++ Program) BGI and UGR coincides with each other very well, even better than Radiance. So the methodology in BGI and UGR calculation for HDR images is good. 5.2.3 GlareIndices (C++ Program) Luminance Histogram Method GlareIndices (C++ Program) has the function of Luminance Histogram Method which is not included in Radiance. But as discussed in Chapter 4, Luminance Histogram Method involves operator’s judgment. And the results are not coinciding with other glare indices like BGI and UGR very well. It seems the results are over exaggerated because of the expanded luminance range and precision and the operator’s subject glare definition/operation. 5.2.4 Luminance Histogram Method As indicated by the tests in Chapter 4, the current criterion may not be plausible for HDR. Luminance Histogram Method criterion for high dynamic luminance range may be developed based on further tests. But based on the experiments results, the HDR Luminance Histogram Method criterion may be adaptive to the actual luminance range within the scene. The criterion, as mentioned in Chapter 4 may not be some fixed numbers. They may be different when applied with different luminance range. On the 147 other hand, Luminance Histogram Method does not consider light sources (possible glare sources) positions. This may cause certain error when the light sources are not in the center of view and the position index is crucial in calculation. So it is found from the experiments that Luminance Histogram Method is better to be used in limited field of view like a task area glare analysis when light sources are limited in a small area (at the center of view). 148 Chapter 6: Future Work Future work should cover several directions as follows: 1. Further tests to find out what are the limits of light levels for the HDR image luminance mapping. This may explain the inflexion in Chapter 3. 2. Further tests to demonstrate if Bracket has consistent better luminance mapping performance than Photosphere. More tests of different software to find out the most accurate HDR image creating software for luminance mapping purpose. 3. Further tests of different camera settings to check how much the influence of camera setting is going to affect luminance mapping accuracy. 4. Further tests are needed to verify the calculated glare indices BGI and UGR from normal lens HDR images. Are they able to reflect or predict the human eyes glare experience in the real world experiment? This requires human psychological survey over a reasonable amount of people within precisely controlled experiment environment. 5. Further tests are needed to develop the application of Luminance Histogram Method in high dynamic range images. The high dynamic range glare criterion 149 needs to be developed based on human psychological survey over a reasonable amount of people within precisely controlled experiment environment as described in No.4. 6. The GlareIndices (C++ Program) may need further development to read other HDR fortmats. 150 References Halsted, C. P. "Brightness, luminance, and confusion." Information Display, 1993: 9 (3): 21. Illustration of Vision. http://www.phys.ufl.edu/~avery/course/3400/vision/eyetype.gif, n.d. Inanici, Mehlika, and Jim Galvin. "Evaluation of High Dynamic Range Photography as a Luminance Mapping Technique." Lawrence Berkeley National Laboratory. http://www.osti.gov/servlets/purl/841925-QBBn0i/native/. , 2004. Iwata, Toshie, Kenichi Kimura, Masanori Shukuya, and Kyosuke Takano. "Discomfort Caused by Wide-source Glare." Energy and Buildings, n.d.: Pages: 391-398. J. R. Coaton, and A. M. Marsden. Lamps and Lighting. London: Arnold, 1997. Jacobs, Axel. WebHDR. n.d. http://luminance.londonmet.ac.uk/webhdr/. Jacobs, Axel. "High Dynamic Range Imaging and its Application in Building Research" Advances in Building Energy Research, 2007. Jan Wienold, Jens Cristoffersen. "Evaluation method and development of a new glare prediction model for daylight environments with the use of CCD camera." Energy and Buildings, 03 17, 2006: 743-757. John E. Kaufman, and Jack F. Christensen. IES lighting handbook: the standard lighting guide. 5th ed. New York: Illuminating Engineering Society, 1978. Melo, Odalea. "An Assessment on the Limitations of the Unified Glare Rating UGR System." n.d. Philips. www.usa.lighting.philips.com. n.d. "Radiance ." n.d. http://radsite.lbl.gov/radiance/framew.html. Schiler, Marc. Toward a Definition of Glare: Can Qualitative Issues Be Quantified? Paris, France: 2nd EAAR-ARCC Conference on Architectural Research, 2000. 151 Stein, Benjamin. Mechanical and electrical equipment for buildings. Hoboken, N.J.: Wiley, 2006. Vision Field. http://www.ssc.education.ed.ac.uk/courses/vi&multi/vmay06c.html, n.d. Ward, Gregory J. "RADIANCE Visual Comfort Calculation." 4 6, 1992. http://radsite.lbl.gov/radiance/refer/Notes/glare.html. Wiki. Bayer Arrangement of Color Filter. http://en.wikipedia.org/wiki/File:Bayer_pattern_on_sensor.svg, n.d. Wiki. Camera CCD Sensor. en.wikipedia.org/wiki/Digital_camera, n.d. Wiki. Electromagnetic Spectrum of Light. http://en.wikipedia.org/wiki/File:EMspectrum.svg, n.d. Wiki. Foveon X3 Sensor. http://en.wikipedia.org/wiki/Foveon_X3_sensor, n.d. Wiki. Nikon's Dichroic Color Seperation Patent Drawing. NikonDichroicPatent.png, n.d. Wiki. Visible Spectrum. http://en.wikipedia.org/wiki/Visible_spectrum, n.d. Wiki. Young's Diffraction of Light. http://en.wikipedia.org/wiki/File:YoungDiffraction.png, n.d. 152 Appendix: Codes Incompleted C++ code which reads 4-byte RGBE file is written by Bruce Walter (bjw@graphics.cornell.edu) 5/26/95 and based on code written by Greg Ward. /* default error routine. change this to change error handling */ static int rgbe_error(int rgbe_error_code, char *msg) { switch (rgbe_error_code) { case rgbe_read_error: perror("RGBE read error"); break; case rgbe_write_error: perror("RGBE write error"); break; case rgbe_format_error: 153 fprintf(stderr,"RGBE bad file format: %s\n",msg); break; default: case rgbe_memory_error: fprintf(stderr,"RGBE error: %s\n",msg); } return RGBE_RETURN_FAILURE; } /* standard conversion from float pixels to rgbe pixels */ /* note: you can remove the "inline"s if your compiler complains about it */ static INLINE void float2rgbe(unsigned char rgbe[4], float red, float green, float blue) { float v; 154 int e; v = red; if (green > v) v = green; if (blue > v) v = blue; if (v < 1e-32) { rgbe[0] = rgbe[1] = rgbe[2] = rgbe[3] = 0; } else { v = frexp(v,&e) * 256.0/v; rgbe[0] = (unsigned char) (red * v); rgbe[1] = (unsigned char) (green * v); rgbe[2] = (unsigned char) (blue * v); rgbe[3] = (unsigned char) (e + 128); } 155 } rgbe2float(float *red, float *green, float *blue, unsigned char rgbe[4]) { double f; if (rgbe[3]) { /*nonzero pixel*/ f = ldexp(1.0,rgbe[3]-(int)(128+8)); //124 *red = (rgbe[0]+0.5) *f; *green =( rgbe[1]+0.5) * f; *blue = (rgbe[2]+0.5) * f; } else *red = *green = *blue = 0.0; } int RGBE_ReadHeader(FILE *fp, int *width, int *height, rgbe_header_info *info) 156 { char buf[128]; // char int found_format; float tempf; int i; found_format = 0; if (info) { info->valid = 0; info->programtype[0] = 0; info->gamma = info->exposure = 1.0; } if (fgets(buf,sizeof(buf)/sizeof(buf[0]),fp) == NULL) return rgbe_error(rgbe_read_error,NULL); if ((buf[0] != '#')||(buf[1] != '?')) { 157 /* if you want to require the magic token then uncomment the next line */ /*return rgbe_error(rgbe_format_error,"bad initial token"); */ } else if (info) { info->valid |= RGBE_VALID_PROGRAMTYPE; for(i=0;i<sizeof(info->programtype)-1;i++) { if ((buf[i+2] == 0) || isspace(buf[i+2])) break; info->programtype[i] = buf[i+2]; } info->programtype[i] = 0; if (fgets(buf,sizeof(buf)/sizeof(buf[0]),fp) == 0) return rgbe_error(rgbe_read_error,NULL); } for(;;) { 158 if ((buf[0] == 0)||(buf[0] == '\n')) return rgbe_error(rgbe_format_error,"no FORMAT specifier found"); else if (strcmp(buf,"FORMAT=32-bit_rle_rgbe\n") == 0) break; /* format found so break out of loop */ else if (info && (sscanf(buf,"GAMMA=%g",&tempf) == 1)) { info->gamma = (float)tempf; info->valid |= RGBE_VALID_GAMMA; } else if (info && (sscanf(buf,"EXPOSURE=%g",&tempf) == 1)) { info->exposure = tempf; info->valid |= RGBE_VALID_EXPOSURE; } if (fgets(buf,sizeof(buf)/sizeof(buf[0]),fp) == 0) return rgbe_error(rgbe_read_error,NULL); } 159 if (fgets(buf,sizeof(buf)/sizeof(buf[0]),fp) == 0) return rgbe_error(rgbe_read_error,NULL); if (strcmp(buf,"\n") != 0) return rgbe_error(rgbe_format_error, "missing blank line after FORMAT specifier"); if (fgets(buf,sizeof(buf)/sizeof(buf[0]),fp) == 0) return rgbe_error(rgbe_read_error,NULL); if (sscanf(buf,"-Y %d +X %d",height,width) < 2) return rgbe_error(rgbe_format_error,"missing image size specifier"); return RGBE_RETURN_SUCCESS; } Incomplete GlareIndice calculation code by the author: rgbe_header_info head; RGBE_ReadHeader(file, &width, &height, &head); 160 float (*data)[3] = new float[width * height][3]; RGBE_ReadPixels(file, data[0], width * height); fclose(file); luminance= new float [width*height]; for(int j = 0; j < height; j++) { for(int k = 0; k < width; k++) { int index = j * width + k; luminance[index]= data[index][0] * 38 + data[index][1] * 128 + data[index][2] * 13; luminance[index]/=head.exposure; float showluminance; 161 showluminance= luminance[index]; } } void FindGlare(float *a, int *b, int index, int zoneNum, bool *ifChange, int width, int height, float ave) { if(a[index]>= (7*ave) &&b[index]==0) { b[index]=zoneNum; *ifChange = true; if(index> width){ FindGlare(a,b,index-width,zoneNum,ifChange, width, height,ave); // find up } 162 if((index+1)% width!=0){ FindGlare(a,b,index+1,zoneNum,ifChange,width,height,ave); // find right } if(index% width!=0){ FindGlare(a,b, index-1,zoneNum,ifChange,width,height,ave); // find left } if(index<(width* height-width)){ FindGlare(a,b,index+width,zoneNum,ifChange,width,height,ave); // find down } } }
Abstract (if available)
Abstract
People’s perception of glare is primarily based on two factors: absolute brightness and contrast. One possible glare analysis method is using empirical formulas to predict people’s feeling of the possible magnitude of un-comfortableness. High Dynamic Range image (HDR) is a type of processed digital image that contains much higher dynamic range information which could potentially be used a luminance mapping tool. In the thesis, author tested several factors which could affect the accuracy of luminance mapping by HDR image and what the extent of influence they are. By using HDR images as luminance maps, the author proposed a C++ program GlareIndices to calculate three glare indices namely Unified Glare Rating, BRS glare equation and Luminance Histogram Ratio to predict glare possibility in an existing scene without involving human subjects.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Absolute glare and relative glare factors: predicting and quantifying levels of interior glare and exterior glare caused by sunlight and daylight
PDF
Post-treatment analysis of the glare remediation of the Walt Disney Concert Hall
PDF
Solar thermal cooling and heating: a year-round thermal comfort strategy using a hybrid solar absorption chiller and hydronic heating scheme
PDF
Guidelines to airport design: accounting for glare from buildings during takeoff and landing – an LAX case study
PDF
Harvesting energy with solar panels and adaptive shading for building skins: A case study of an office building in Saudi Arabia
PDF
Blind to light loss: evaluating light loss through commercial building facades as a contribution to urban light pollution
PDF
Dynamic shading and glazing technologies: improve energy, visual, and thermal performance
PDF
Human-building integration: Investigation of human eye pupil sizes as a measure of visual sensation in the workstation environment
PDF
Disaster relief transitional emergency shelter: environmental and structural analysis of two prefab modular emergency shelters for three different Californian climate zones
PDF
Comparison of four different types of prefab homes and a site-built home
PDF
Comparison between tooth mesiodistal angulation measurements from constructed panoramic images and three-dimensional volumetric images
PDF
Exploration for the prediction of thermal comfort & sensation with application of building HVAC automation
PDF
Evaluation and development of solar control screens: using daylight simulation to improve the performance of facade solar control screens
PDF
Impacts of building performance on occupants' work productivity: a post occupancy evaluation study
PDF
Tilted glazing: angle-dependence of direct solar heat gain and form-refining of complex facades
PDF
Building bridges: filling gaps between BIM and other tools in mechanical design
PDF
Energy use intensity estimation method based on building façade features by using regression models
PDF
Kinetic facades as environmental control systems: using kinetic facades to increase energy efficiency and building performance in office buildings
PDF
Energy efficient buildings: a method of probabilistic risk assessment using building energy simulation
PDF
An analysis of underwater habitats a development of the outline for aquatectural graphic standards
Asset Metadata
Creator
Yin, Hanshu
(author)
Core Title
Glare studies: Comparison of three glare indices, HDR imaging and measured values
School
School of Architecture
Degree
Master of Building Science
Degree Program
Building Science
Publication Date
05/03/2011
Defense Date
04/01/2011
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
glare,glare indices,HDR,OAI-PMH Harvest
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Schiler, Marc E. (
committee chair
), Suk, Jae Yong (
committee member
), Woll, Edwin (
committee member
)
Creator Email
ffcyhs@hotmail.com,hanshuyi@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-m3878
Unique identifier
UC1156685
Identifier
etd-Yin-4582 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-472568 (legacy record id),usctheses-m3878 (legacy record id)
Legacy Identifier
etd-Yin-4582.pdf
Dmrecord
472568
Document Type
Thesis
Rights
Yin, Hanshu
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Repository Name
Libraries, University of Southern California
Repository Location
Los Angeles, California
Repository Email
cisadmin@lib.usc.edu
Tags
glare
glare indices
HDR