1	Graphics File Format and Data Compression Techniques (take 2) Week 7
2	Business Attendance (please sign in)
3	Tonight Homework 6 Lecture Midterm Review If we finish early we leave early
4	No homework this week Please complete all reading Review for the exam Schedule office hours to clear-up any questions
5	Midterm Next week: Monday November 1^st We will have the exam followed by lecture 1 hour to complete exam Break Resume at 8 for brief lecture Exam will be closed book No calculator or computer
6	Homework 6 What did you think of the HW?
7	Note on Homeworks I have examined Most work Some: Only go RGB -> HSL Don’t display the resulting color
8	C# Surprising number of people requested C# OK, but really need to document code
9	Note on App builders Very nice to quickly build UI However: I don’t intend to hunt for your code between a dozen files full of auto generated code and content Use a framework if you wish but Condense and comment the results
10	Interactive RGB óHSL insights People think of color in terms of Hue, Saturation/Tint, and Luminance/Brightness People rarely consider the relative amounts of Red, Green, and Blue in a color
11	Interactive RGB óHSL insights Variable amounts of any three colors can define the same color Providing the color defined is within all gamuts
12	Engineering insights Easiest to program given an internal representation of color Some feeling of data representation Degrees, percentages, etc. as bits
13	Not the original assignment Wanted you to deal with: Mapping colors between color spaces Some colors are not representable Specifically Mapping Computer’s RGB space into NTSC video
14	Illegal Colors Need Phosphor Chromaticity: NTSC ‘theoretical’ Common NTSC TV Common RGB Monitors
15	Brightest is best Brightest TV sells the best Brighter blue phosphor Eye insensitive to blue So only 5% brighter Whopping 9300K white (Compare to 6500K sunlight) At night ever walk past a window with a TV flickering?
16	8 Laws of Color Matching Any color may be matched by mixing at most 3 lights Luminance of a color = sum of the luminance of its components Humans cannot distinguish components of a color match Metameric match is a pair indistinguishable colors with different spectra
17	8 Laws of Color Matching cont. A color match works over a range of luminances If 2 colors match their mixtures will match If 2 colors match and matching colors are subtracted from each the results match If C₁ matches C₂, and C₂ matches C₃, then C₁ matches C₃
18	CIE Commission Internationale de L’Eclairage CIE tristimulus color coords XYZ
19	Color Matching CRT Monitors Theory: Given CIE values of the 3 CRT monitor phosphors And the CIE values for the RGB the monitor uses Original colors can be matched
20	Converting xyY ó XYZ xyY: x,y chromaticity, Y luminance
21	Color Matching CRT Monitors Problem: Monitor specifications are rarely made available by the manufacturer
22	For most monitors xyY=(0.285, 0.293, 1) XYZ=(0.973, 1, 1.440) Corresponds to 9300K white point!
23	CIE and file formats Why isn’t CIE used in most file formats? Positives: Can encode all visible colors Colors are universal Not device specific
24	CIE and file formats Issues: Represents many colors not representable by hardware Larger color space requires more precision (bits to encode at same resolution) Most hardware is non-linear and poorly characterized PNG and proprietary formats encode RGB But provide additional information about the capture device
25	Color matching monitors Matching device specific RGB colors (on, say, 2 video monitors) Accomplished by 1^st converting into a common device independent space (CIE space is convenient) Then converting back into the 2^nd device’s RGB space using an inverse matrix
26	Color conversion using matrices This is a straightforward matrix operation See Miano p197 for details NOTE: Normalize all values to 0-1 ranges Resultings coordinates <0 or >1 are not representable by the device Called illegal colors
27	Illegal colors Two choices for dealing with illegal colors ‘Limit’ components of illegal colors Changes the color Changes the mix of colors Or compress the color space of the image so all colors ‘fit’ This changes the range of colors Preserves the color ‘dynamics’
28	MacAdam Ellipses MacAdam ellipses describe regions of CIE space that humans perceive as indistinguishable (Just noticeable difference) Initially used human subjects on discrimination tests Unsurprisingly human perception of CIE space isn’t uniform
29	MacAdam Ellipses Red ellipses -> measured Blue ellipses -> modeled Based on non-linearities in the cones From: Photoreceptor non-linearities can account for the MacAdam ellipses
30	Uniform Chromaticity Attempts to make color distance = difference CIE Uniform chromaticity Scale (UCS) uvY UVW Lab
31	Gamma Most systems have non-linear response Consider: Film RGB monitors Photoreceptors in our Eyes Gamma represents the response curve
32	Gamma curves Gamma of 1: linear Gamma < 1: low contrast Provides more latitude in dark region Gamma > 1: high contrast CRT monitors: typically 2.0-3.0
33	Gamma in HW/SW Most systems don’t provide HW gamma correction Correspondingly use SW to change the RGB values for pixels in an image Gamma < 1 Assigns more values to darker colors (think palate)
34	Gamma curves 0.5 Curve: 2.0 curve:
35	Gamma in HW High end systems Like the SGI systems I build Use an 8 bit gamma palate per RGB channel Fast Non-destructive
36	Gamma issues Gamma more than a function of hardware Gamma also changes based on lighting Gamma can be approximated Using specially prepared image Trading spatial illumination ½ pixels black, ½ pixel max Should equal a region of ½ intensity
37	Gamma Example
38	Gamma Challenges Important to calibrate gamma when creating images Unfortunately PC, Web Content, etc. Authored expecting non-linear gamma Typical PC and Mac gammas differ Content created for 1 Looks bad the another Looks bad on properly calibrated display
39	To fix color problem we need Properly characterized Display devices Video output Image formats in device independent color space Display programs that map device independent to device dependent color Hardware acceleration (so there isn’t a ‘penalty’
40	Exam Review: Methodology Don’t expect memorization Other than important constants and concepts Pays to recall if relationships are Linear Geometric Asymptotic Inverse Etc.
41	Exam Review: Questions I don’t like multiple choice/fill in the blank Prefer questions that require thought and demonstrate understanding Expect to: Work mathematical problems Graph Write simple programs Justify a conclusion
42	Topics, Be sure to: Review all lecture notes Will be emphasizing what was covered in class Complete all reading Will expect understanding of this material if not memorization Be able to reproduce any homework
43	Example Topics Data representation Analog vs Digital Sampling Quantization Human Perception Just noticeable difference Sensors
44	Example Topics Information theory Entropy Signal/Noise Entropy Coding Huffman coding Dictionary coding LZ family
45	Example Topics Color Color Perception Tristimulus color Color matching JND Color Representation Color Spaces