Opponent Color Spaces

Transcription

1 C. A. Bouman: Digital Image Processing - January 8, Opponent Color Spaces Perception of color is usually not best represented in RGB. A better model of HVS is the so-call opponent color model Opponent color space has three components: O 1 is luminance component O 2 is the red-green channel O 2 = G R O 3 is the blue-yellow channel Comments: O 3 = B Y = B (R+G) People don t perceive redish-greens, or bluish-yellows. As we discussed, O 1 has a bandpass CSF. O 2 and O 3 have low pass CSF s with lower frequency cut-off.

2 contrast sensitivity C. A. Bouman: Digital Image Processing - January 8, Opponent Channel Contrast Sensitivity Functions (CSF) Typical CSF functions looks like the following. Luminance CSF Red-Green CSF Blue-Yellow CSF cycles per degree

3 C. A. Bouman: Digital Image Processing - January 8, Consequences of Opponent Channel CSF Luminance channel is Bandpass function Wide band width high spatial resolution. Low frequency cut-off insensitive to average luminance level. Chrominance channels are Lowpass function Lower band width low spatial resolution. Low pass sensitive to absolute chromaticity (hue and saturation).

4 C. A. Bouman: Digital Image Processing - January 8, Some Practical Consequences of Opponent Color Spaces Analog video has less bandwidth in I and Q channels. Chrominance components are typically subsampled 2-to- 1 in image compression applications. Black text on white paper is easy to read. (couples to O 1 ) Yellow text on white paper is difficult to read. (couples to O 3 ) Blue text on black background is difficult to read. (couples to O 3 ) Color variations that do not change O 1 are called isoluminant. Hue refers to angle of color vector in (O 2,O 3 ) space. Saturation refers to magnitude of color vector in (O 2,O 3 ) space.

5 C. A. Bouman: Digital Image Processing - January 8, Opponent Color Space of Wandell First define the LMS color system which is approximately given by L M S = The opponent color space transform is then 1 O 1 O 2 O 3 = We many use these two transforms together with the transform from srgb to XYZ to compute the following transform. O 1 O 2 O 3 = L M S Comments: O 1 is luminance component O 2 is referred to as the red-green channel (G-R) O 3 is referred to as the blue-yellow channel (B-Y) Also see the work of Mullen 85 2 and associated color transforms. 3 X Y Z sr sg sb 1 B. A. Wandell, Foundations of Vision, Sinauer Associates, Inc., Sunderland MA, K. T. Mullen, The contrast sensitivity of human color vision to red-green and blue-yellow chromatic gratings, J. Physiol., vol. 359, pp , B. W. Kolpatzik and C. A. Bouman, Optimized Error Diffusion for Image Display, Journal of Electronic Imaging, vol. 1, no. 3, pp , July 1992.

6 C. A. Bouman: Digital Image Processing - January 8, Paradox? Why is blue text on yellow paper easy to read?? Shouldn t this be hard to read since it stimulates the yellowblue color channel?

7 C. A. Bouman: Digital Image Processing - January 8, Better Understanding Opponent Color Spaces The XYZ to opponent color transformation is: O O 2 = O = v y v gr v by What arev y, v gr, and v by? X Y Z They are row vectors in the XYZ color space. v gr is a vector point from red to green v by is a vector point from yellow to blue They are not orthogonal! X Y Z

8 C. A. Bouman: Digital Image Processing - January 8, Plots ofv y,v gr, and v by 0.9 Opponent Color Directions of Color Matching Functions 0.8 Green Yellow 0.4 Green Red 0.3 D65 white point Red 0.2 Yellow Blue 0.1 blue

9 C. A. Bouman: Digital Image Processing - January 8, Answer to Paradox Since v y, v gr, and v by are not orthogonal v y v gr [ vyv t grvby] t t identity matrix v by Blue text on yellow background produces and stimulus in thev by space. O 1 v y O 2 = v gr vby t = O 3 v by This stimulus is not isoluminant! Blue is much darker than yellow.

10 C. A. Bouman: Digital Image Processing - January 8, Basis Vectors for Opponent Color Spaces The transformation from opponent color space to XYZ is: X O 1 Y = O 2 Z O 3 O 1 = [c y c gr c by ] O 2 O 3 What arec y,c gr, and c by? They are column vectors in XYZ space. c gr is a vector which has no luminance component. c by is a vector which has no luminance component. They are orthogonal to the vectorsv y, v gr, and v by.

11 C. A. Bouman: Digital Image Processing - January 8, Plots ofc y, c gr, and c by 0.9 Opponent Color Directions of Color Matching Functions 0.8 Green Yellow 0.4 Green Red 0.3 Red Yellow Blue blue

12 C. A. Bouman: Digital Image Processing - January 8, Interpretation of Basis Vectors Sincec y,c gr, andc by are orthogonal tov y,v gr, andv by, we have v y v gr [c y c gr c by ] = v by Therefore, we have that O 1 v y O 2 = v gr c by O 3 v by = = 0 1 So, c by is an isoluminant color variation Something like a bright saturated blue on a dark red.

13 C. A. Bouman: Digital Image Processing - January 8, Solution to Paradox Why is blue text on yellow paper is easy to read?? Solution: The blue-yellow combination generates the inputv by. This input vector stimulates all three opponent channels because it is not orthogonal to c y, c gr, and c by. In particular, it strongly stimulates c y because it is not iso-luminant.