Color in Image Processing

Reference: Digital Image Processing (ECE 6258) from Prof. Ghassan Alregib

---

• Pseudo-color is used when representing the targeted object to be shown more clearly.
• Transform or scale the RGB value for representing pseudo-color.

• Color is determined by the source of the light, texture of the object, and the perception method.
• Even identical camera capture the color in very different way.

 

• Human can distinguish the difference between cats and dogs.
• But classifying problem of cats and dogs are difficult for computer, because understanding the color difference is difficult.
• Understanding and processing color could be really good descriptor for vision camera.

 

• Achromatic color has no hues, no saturation, only has its intensity(brightness).
• Visible light spans from 400 to 700nm -> chromatic light spans from 400 to 700nm.

• Human eyes has 2 kinds of cells in order to perceive the light.
• 3 kinds of cone cells(Phtopic vision) perceive the color.
  • L(Red): 63% distribution
  • M(Green): 31% distribution
  • S(Blue): 6% distribution
• Rod cells(Scotopic vision) could not differentiate the color.

 

• There are some colors that could not be made up by mixing the primary colors. **(?) (misconception)
• Every three colors, if their wavelengths are far enough apart, can serve as primary colors.
• Primary colors are independent. You cannot mix any other colors to make these.

 

• Hue(rotation) indicates the color difference.
• The boundary of the circle indicates the pure color.
• Degree of saturation is inversely proportional to the amount of white added.
• That is, saturation indicates "How far I go from the origin".
• Hue and Saturation, which is chromaticity, is important descriptor of color.

 

• Color is characterized by intensity(brightness) and chromaticity.
• For example, white and gray have the same chromaticity but different brightness.
• CIE defines three color function x(lambda), y(lambda), and z(lambda) by the cone cells' reaction to the tristimulus values X,Y, and Z each. Here, lambda denotes to wavelength.
• Since Y indicates the brightness, the chromaticity could be represented by two values, x, and y.
• The color space represented by x,y, and Y is called CIE xyY color space.
• Here, X and Z could be obtained by
  • X = x*Y/y
  • Z = (1-x-y)*Y/y

 

the colors that shown in chromaticity diagram is the colors that we can actually see.

- boundary= pure color(no other color componenet)

if white added, low saturation(look below)

- J&D

- start and stop when I think the color is different

green region: 3mm

red: 1mm

nonlinearlity

so weighted 

human beings (how do you measure the J&D)

 

+LAB (correlated

L: nominance

A/B: hues, 

HSI much less correlation.

So many use HSI LAB because of this issue.