CDS 6334 - Visual Image Processing

Lecture 6: Colour

1. What is Colour?

Colour is the result of the interaction between physical light and the human visual system.
Colour is a psychological perception, not a physical property of an object.
🧠 Remember:

Colour exists in the brain, not the object.

2. Newton's Discovery

In 1666, Isaac Newton showed that white sunlight can be separated into a spectrum of colours using a prism.
White light contains many wavelengths of visible light.

3. Colour and Wavelength

Visible light occupies approximately 400 nm to 700 nm of the electromagnetic spectrum.
Colour Approximate Wavelength
Violet ~400 nm
Green 500–570 nm
Red ~700 nm

4. Human Colour Perception

The retina contains light-sensitive photoreceptors called rods and cones.
Photoreceptor Function
Rods Brightness and night vision
Cones Colour vision

5. Types of Cones

Cone Type Most Sensitive To
S Cone Blue
M Cone Green
L Cone Red
The brain combines signals from these three cone types to perceive colour.
Exam Fact:
Human colour vision is trichromatic.

6. Additive Colour Mixing

Colours are created by adding coloured light together.
Red + Green = Yellow
Red + Blue = Magenta
Green + Blue = Cyan
All three primary lights combined produce white.
🧠 Additive:
Black → White

7. Subtractive Colour Mixing

Colours are produced by removing wavelengths from white light.
Cyan + Magenta + Yellow → Black
Used in printing and pigments.
🧠 Subtractive:
White → Black

8. Problems with Colour Images

Important:
Lighting greatly affects recorded RGB values.

9. Colour Spaces

A colour space is a coordinate system used to represent colours numerically.
Each colour corresponds to a point within the colour space.

10. Common Colour Spaces

Colour Space Main Use
RGB Displays
CMY / CMYK Printing
HSV Human-friendly colour representation
YCbCr Digital video
CIE LAB Perceptual colour measurement

11. RGB Colour Space

RGB represents colour using Red, Green and Blue components.
It is an additive colour model commonly used in displays.
🧠 RGB = Screens

12. RGB Colour Cube

RGB colours form a cube in 3D coordinate space.

13. CMY and CMYK

CMY uses Cyan, Magenta and Yellow as primary colours.
Conversion:
C = 1 − R
M = 1 − G
Y = 1 − B
CMYK adds black ink (K) for true black printing.

14. Colour Gamut

Gamut refers to the range of colours represented by a colour model.
RGB generally has a larger gamut than CMY.
🧠 Not all screen colours can be printed.

15. HSV Colour Space

HSV represents colours using Hue, Saturation and Value.
Component Meaning
Hue Dominant colour
Saturation Colour purity
Value Brightness

16. Why HSV is Popular

HSV is more intuitive and closer to human colour perception than RGB.
Useful for colour segmentation and object detection.
🧠 HSV = Human-Friendly RGB

17. YCbCr Colour Space

YCbCr separates brightness information from colour information.
Component Description
Y Luminance
Cb Blue difference
Cr Red difference
Commonly used in digital video systems.

18. Colour Difference

Colour difference measures how different two colours are.
Euclidean distance is the simplest colour distance measure.
Colour Difference:
Distance between colour coordinates.

19. Perceptually Uniform Colour Spaces

Equal distances should correspond to equal perceived colour differences.
RGB, HSV and YCbCr are not perceptually uniform.
Better Choices:
CIE LAB and CIE LUV

20. CIE LAB and CIE LUV

Designed so colour distances better match human perception.
Component Meaning
L* Lightness
a* Green ↔ Red axis
b* Blue ↔ Yellow axis

21. Colour-Based Image Retrieval (CBIR)

Images can be retrieved based on colour similarity.
Colour histograms are commonly used as image descriptors.
🧠 Similar Histogram = Similar Colour Distribution

22. Colour Histograms

A colour histogram describes the distribution of colours within an image.
Independent of image position, rotation and scale.

23. Colour-Based Skin Detection

Human skin tones occupy specific regions in colour spaces such as RGB, HSV and YCbCr.
Frequently used in face detection systems.

24. Colour Harmonization

Adjust colours within an image to create visually pleasing combinations.
Used in photography, graphic design and image editing.

25. Colourization

Automatically adds colour to grayscale images.
Modern colourization systems commonly use deep learning.

26. Final Exam Summary

Most Important Points

  • Colour: Interaction between light and visual perception.
  • Visible Spectrum: Approximately 400–700 nm.
  • Rods: Brightness vision.
  • Cones: Colour vision.
  • Three Cone Types: S, M and L.
  • RGB: Additive colour model.
  • CMY/CMYK: Subtractive colour model.
  • HSV: Hue, Saturation, Value.
  • YCbCr: Luminance + Colour Difference.
  • Colour Gamut: Range of representable colours.
  • RGB is not perceptually uniform.
  • CIE LAB and CIE LUV: Perceptually meaningful colour spaces.
  • CBIR: Uses colour histograms for image retrieval.
  • Applications: Skin detection, harmonization and colourization.