1. Image Perception and Color Space

2. Human Eye Photoreceptors: 1. Rods more sensitive, for low light
2. Cones less sensitive, for day light (So, human can’t perceive color in low light)
100M rods vs. 6.5M cones. -> 16:1 -> chrominance down-sampling can be 4:1 in each dimension
The Fovea (center of retina) All the cone sensors of the retina are concentrated in this area. It is a small pale spot, about the size of a pinhead and, on our big picture, is in the darkened area just to the right of the optic disc - it has been electronically enhanced in this image. It is totally responsible for our colour vision and our critical vision

3. Simultaneous Contrast
luminance vs. brightness
Contrast
Weber’s Law:
If background is f, than foreground of (f+delta f) gives you the same contract

4. Mach Band Lateral inhibition Eye sees contrast, not absolute luminance
So, when the neighbor is bright, the area looks darker
Can be modeled as a sinc low pass filter for the impulse response of visual system

5. Herman Grid
Can be explain by later inhibition (when there are white above/below/left/right, it looks dark)
When you focus on one intersection, the other intersection turn dark, which proves that lateral inhibition is more significant in the peripheral

6. Blind Spot Test Where the optical nerve is
Ask the viewer to use right eye, focus on the cross

7. Blind Spot Test Where the optical nerve is
Ask the viewer to use right eye, focus on the cross

8. Blind Spot Test Where the optical nerve is
Ask the viewer to use right eye, focus on the cross

9. Blind Spot Test Where the optical nerve is
Ask the viewer to use right eye, focus on the cross

10. Color Representation

11. Visual Spectrum
Blue nm
Green nm
Red 700 nm

12. Perceptual Representation (HSV)

13. CIE Color Chart CIE XYZ Not uniform chromaticity scale (UCS)
CIE: Intl Committee on Color Standards
CIE XYZ

14. RGB primary colors
CMY secondary colors

15. Lower right: saturated colors, farthest from the line connecting black and white
Upper right: pure red of different intensity/luminance
The upper left is similar to CIE XYZ

16. YUV Color Space Y is luminance of a color
Y = 0.299*R’ *G’ *B’
U and V are color differences
U = 0.492*(B’-Y)
V = 0.877*(R’-Y)
This simplifies recovery of R’,G’,B’
R’ = Y V
G’ = Y U V
B’ = Y U
R’,G’,B’ here are same as R_N,G_N,B_N in Jain’s

17. YIQ Color Space
Rotate UV vectors by 30o U V I Q

18. YIQ Color Space I = 0.736*(R’-Y) - 0.268(B’-Y)
Q = 0.478*(R’-Y) *(B’-Y)
Recovery of R,G,B
R’ = Y *I *Q
G’ = Y *I *Q
B’ = Y *I *Q

19. YCbCr Scaled and offset version of YUV Range of signals
Suitable for digital images/video

20. YCbCr Color Space Begin by calculating R-Y and B-Y vectors
B’-Y = *R’ *G’ *B’
R’-Y = 0.701*R’ *G *B’
The difference signals have ranges
B’-Y (-.866 to .866)
R’-Y (-.701 to .701)
Scale to range (-.5 to .5) to give same range as Y (0 to 1)
Cb = -.169*R’ *G’ *B’ = .564(B’-Y)
Cr = .500*R’ *G’ *B’ = .713(R’-Y)

21. YCbCr Color Space
For computer representation, Scale and offset these values to keep in range 16 to 240 (Cb and Cr) or 16 to 235 (Y)
Cb = 224*Cb = (B’-Y) + 128
Cr = 224*Cr = (B’-Y) + 128
Y = 219*Y + 16
Finally,

22. CMYK Color Space Subtractive color space
Most common use is for printers
K (black) is added for efficiency and consistency
White cannot be generated w/o white paper
Example: Cyan represents green and blue, by adding cyan we subtract the color red from sum.

23. Perception? Illusion?

24. Masking Effect
Quantization step size = 32

25. Masking Effect
Quantization step size = 32
But less visible artifacts

26. Necker Cube
Tiring effect: the more you look at it, the more unstable it appears to jump between different interpretation

27. How many? 6 or 7?

28. 3 different interpretations

29. Which is longer?
Explanation: eye trends to compare closer parts of the two objects

30. Poggendorff Illusion

31. Ponzo Illusion
The gray lines produce the distance effect, human eyes magnify object that is farther
The one that “fills” the space looks longer

32. Adaptation
Eyes are sensitive to changes, but get used to stead states

33. Adaptation

34. Adaptation

35. Adaptation

36. Are they straight lines?

37. Which is longer?
The horizontal is “broken” by the vertical, so looks shorter

38. Stereo Vision

39. Autostereogram

40. What do you see?

41. What do you see?