A World of Color. Session 4 Color Spaces. OLLI at Illinois Spring D. H. Tracy

Transcription

1 A World of Color Session 4 Color Spaces OLLI at Illinois Spring 2018 D. H. Tracy

2 Course Outline 1. Overview, History and Spectra 2. Nature and Sources of Light 3. Eyes and Color Vision 4. Color Spaces and Color Mixing 5. Origins of Colors of Things 6. Visual Perception: Neurons and the Brain 7. Color Recording and Reproduction, Media 8. Color in Arts and Culture, Emotive Aspects 2/15/2018 World of Color 4 2

3 Session 4 Outline Color Spaces Review of vision process How cones respond to colors Color Models Traditional Models The CIE Color Model Models of Color Blindness The role of luminance Photometry How our color perception is measured 2/15/2018 World of Color 4 3

4 Color Perception Train Source Brain Optic Nerve Object Medium Eye 2/15/2018 World of Color 4 4

5 Review of Color Vision in the Retina 2/15/2018 World of Color 4 5

6 (~1000) Rods and Cones ~25μm (~2 μm dia) ic.ucsc.edu slide show Anatomy of the human eye Bruce Bridgeman, UC Santa Cruz ~ 1 Hair UCSC Vision Course 2/15/2018 World of Color 4 6

7 2 Isomers of Retinal Trans- Retinal PHOTON Cis Retinal 2/15/2018 World of Color 4 7

8 Opsins Members of G protein-coupled receptors class ~ 400 Amino acids long > 600 Million Years old ~ 100K per Disc Trans-membrane Proteins (7 sections) 2/15/2018 World of Color 4 8

9 Opsins in Human Eye Opsin Abbr. Human Variant Location Peak Wave length Rhodopsin Rh Rho = OPN2 Rods 505 nm L-Cone (Red) LWS OPN1 LW Cones (Red) M-Cone (Green) MWS OPN1 MW Cones (Green) S-Cone (Blue) SWS1 OPN1 SW Cones (Blue) Melanopsin MOP OPN4 Ganglion Cells Chromo some 557 X Function 3 Dark Vision (Scoptic) 527 X Photopic Color Vision Circadian Rythm 2/15/2018 World of Color 4 9

10 Relative Sensitivity Typical Human Cone Sensitivities S M L λ 2/15/2018 World of Color 4 10

11 Apples as seen by a Blue-Cone Monochromat 2/15/2018 World of Color 4 11

12 Relative Sensitivity Almost Any Wavelength can Excite M-Cones S M M L λ 2/15/2018 World of Color 4 12

13 Guessing Colors of Light Patches: M-Cones Only M-Cones 2/15/2018 World of Color 4 13

14 Guessing Colors of Light Patches: M-Cones Only M-Cones All Cones 2/15/2018 World of Color 4 14

15 Guessing Colors of Light Patches: M-Cones Only M-Cones All Cones Not Possible 2/15/2018 World of Color 4 15

16 2/15/2018 World of Color 4 16

17 All 3 Cones L-Cones M-Cones S-Cones 2/15/2018 World of Color 4 17

18 All 3 Cones L-Cones M-Cones S-Cones 2/15/2018 World of Color 4 18

19 All Cones L-Cones M-Cones To determine the actual perceived Tristimulus color, we need data from all 3 Cone types: S-Cones S, M and L 2/15/2018 World of Color 4 19

20 Relative Sensitivity Each Cone Type Registers a Signal S M L S M L λ 2/15/2018 World of Color 4 20

21 Organizing Colors 2/15/2018 World of Color 4 21

22 Runge s Farbenkugel (1807) Philipp Otto Runge, Painter ( ) as is known, there are only three colors, yellow, red and blue letter to Goethe, /15/2018 World of Color 4 22

23 Johannes von Goethe (Theory of Colors, 1810) 2/15/2018 World of Color 4 23

24 Munsell Color System Color Trees Albert Munsell, Painter ( ) 2/15/2018 World of Color 4 24

25 Munsell Color Swatches 1600 Color Chips $1029, Amazon Used in World Color Survey 2/15/2018 World of Color 4 25

26 Scientific Development of Tristimulus Theory Tristimulus vs Trichromatic Thomas Young (1802) Conjectured that eye had 3 color channels Herman von Helmholtz (1850 s) Advanced theory, but found experimentally that 3 primary colors are insufficient had doubts James Clerk Maxwell (1860 s) Put theory on a sound mathematical basis. Separated Luminance from Hue and Saturation. Many scientists in late 1800 s and early 20 th century worked out the details & did thousands of human colorimetry measurements CIE 1931 Standard Observer 2/15/2018 World of Color 4 26

27 Obvious Tricolor Model S Cone Signal Each set of cone signals can be plotted as a point in 3-D space O 2/15/2018 World of Color 4 27

28 Obvious Tricolor Model S Cone Signal Doubling each of the 3 cone signals results in the same color, but with double the Luminance. O All such amplified points lie on a line from the origin. 2/15/2018 World of Color 4 28

29 Obvious Tricolor Model S Cone Signal Same color, just twice as bright O Waste of 3D too complex Drop Luminance can use 2D 2/15/2018 World of Color 4 29

30 y CIE* 1931 Color Space (x & y Hue Coordinates) Chromaticity Diagram *CIE = International Commission on Illumination (Commission Internationale de l éclairage) 2/15/2018 World of Color 4 30 x

31 CIE 1931 xy Color Space Important Property: Any two lights can be mixed to produce any color along the line between them, but no other colors! y In this case, a monochromatic teal and a reddish pastel can be mixed to produce sunlight white, e.g. 2/15/2018 World of Color 4 x 31

32 Blackbody Lights plotted on CIE 1931 Chromaticity Diagram 2/15/2018 World of Color 4 32

33 Mixing 3 Lights 3 Monochromatic lights can generate any color within their triangle (Actually, this is true of any 3 lights, even if not monochromatic) 2/15/2018 World of Color 4 33

34 Try to Maximize the Covered Region with 3 Primaries CIE 1931 xy Color Space y Gamut Most extreme 3 lights imaginable, but not enough for the gamut to cover the full chromaticity space! 2/15/2018 World of Color 4 x 34

35 Computer Display Gamuts P3 Gamut CIE 1931 xy Color Space This region cannot be correctly displayed srgb Gamut 2/15/2018 World of Color 4 35

36 CIE 1931 xyz Color Space Chromaticity Diagram S x = y = M L L + M + S M L + M + S 1931 CIE Standard Observer Matching Functions L These are the pseudo cone responses assumed by CIE, consistent with photometry data. They did not know the true cone response curves at the time. Actual z = 1 Cone x y Responses (2016) 2/15/2018 World of Color 4 36

37 CIE 1931 xyz Color Space S M 1931 CIE Standard Observer Matching Functions L As Maxwell knew, it not possible to work out the actual cone responses from color matching data, but we can get into the right family of curves. In many cases, the values go negative! x = L L + M + S red-ness y = M L + M + S Actual z = 1 Cone x y Responses (2016) green-ness (also Lumosity!) blue-ness Each such set which correctly explains the human Standard Observer matching data can be transformed into the infinity of others via a linear transformation (rotation). 2/15/2018 World of Color 4 37

38 CIE 1931 xyz Color Space S M 1931 CIE Standard Observer Matching Functions L As Maxwell knew, it not possible to work out the actual cone responses from color matching data, but we can get into the right family of curves. In many cases, the values go negative! x = L L + M + S red-ness y = M L + M + S Actual z = 1 Cone x y Responses (2016) green-ness (also Lumosity!) blue-ness Each such set which correctly explains the human Standard Observer matching data can be transformed into the infinity of others via a linear transformation (rotation). 2/15/2018 World of Color 4 38

39 Standard Luminosity Curves Gray-scale intensity, without considering color Rods All 3 Cones combined CIE /15/2018 World of Color 4 40

40 Standard Luminosity Curves Gray-scale intensity, without considering color Rods Fun fact: CIE chose this curve (arbitrarily) as their M pseudo-cone curve! All 3 Cones combined CIE /15/2018 World of Color 4 41

41 Since luminosity is fixed and x+y+z=1, this 45 o line is the limit. The yellow and red monochromatic lights are right up against this limit. 2/15/2018 World of Color 4 42

42 Equivalent cone responses 2/15/2018 World of Color 4 43

43 How would a digital camera do it? Typical camera chip channel relative response 2/15/2018 World of Color 4 44

44 How Could Color Perception Go Wrong? S M L Missing Cone λ 2/15/2018 World of Color 4 45

45 Color Blindness (missing Cone) Lines of Confusion Protan (L Cone Missing) Deutan (M Cone Missing) Every color along this line looks exactly the same 2/15/2018 World of Color 4 46

46 Color Blindness (missing Cone) Lines of Confusion Protan (L Cone Missing) Deutan (M Cone Missing) Tritan (S Cone Missing) Every color along this line looks the same 2/15/2018 World of Color 4 47

47 For Deuteranopes, 2-D Chromaticity collapses to 1-D All colors can be arrayed along a single axis. Only hundreds to maybe 1000 hues can be distinguished. 2/15/2018 World of Color 4 48

48 CIE 1931 xy Color Space with MacAdam Ellipses (enlarged 10x for clarity) Observer asked to match lights to a fixed reference (the dot) many times. The results are never exact, all fall within the MacAdam ellipses (here exaggerated 10x). Some parts of chromaticity diagram are harder to distinguish than others. Fixed Luminance (48 cd/m2) Observer Perley G. Nutting 2/15/2018 World of Color 4 49

49 CIE 1976 uv Uniform Chromaticity Scale Derived from CIE 1931 (nonlinear relationship) Hue Closer to Perceptually Uniform: MacAdam ellipses would be circles of almost uniform size. White Point White Point Hue Saturation Purity Colorfulness Luminance (not on plot) Saturation 2/15/2018 World of Color 4 50

50 CIE 1976 uv Uniform Chromaticity Scale Derived from CIE 1931 (nonlinear relationship) Hue Closer to Perceptually Uniform: MacAdam ellipses would be circles of almost uniform size. White Point White Point Hue Saturation Purity Colorfulness Luminance (not on plot) Saturation Not used much since the properties of 2- light mixing along a line, etc., no longer work in this representation. 2/15/2018 World of Color 4 51

51 Metamers CIE 1931 xy Color Space Chromaticity Diagram Mixing 500nm + 600nm can give identical color to 480nm+560nm Very different colors that appear identical are called Metamers 2/15/2018 World of Color 4 52

52 Metamers CIE 1931 xy Color Space Chromaticity Diagram Thousands or Millions of Metamers in this case In a sense we are all profoundly color blind we can t distinguish any of them! 2/15/2018 World of Color 4 53

53 Demo of Color Matching 2/15/2018 World of Color 4 54

54 Demo of Color Matching Results: 1. Very wide disagreement among class members. 2. No perfect match attainable for anyone 2/15/2018 World of Color 4 55

55 Analysis of the Amber Matching Demo CIE 1931 xy Color Space Chromaticity Diagram Green LED Amber LED (Reference Light) Red LED Also, Variations in individuals M and L cones result in disagreement about best Red/Green LED ratio. 2/15/2018 World of Color 4 56

56 Session 4 Outline Color Spaces Review of vision process How cones respond to colors Color Models Traditional Models The CIE Color Model Models of Color Blindness The role of luminance Photometry How our color perception is measured 2/15/2018 World of Color 4 79