Uniform color spaces and natural image statistics

Area of Science:

• Visual Neuroscience
• Color Science
• Computational Vision

Background:

• Efficient visual coding theories often start with natural scene statistics.
• Perceptually uniform color spaces aim to represent equal color differences equally.
• The relationship between color space properties and natural environment statistics is not fully understood.

Purpose of the Study:

• To investigate the type of color environment implied by perceptually uniform color spaces.
• To explore how natural scene statistics relate to color representations in the visual system.
• To determine if environmental color biases influence chromatic sensitivity.

Main Methods:

• Analysis of color distributions within perceptually uniform color spaces (Munsell, CIE L(*)u(*)v(*), CIE L(*)a(*)b(*)).
• Projection of these distributions into cone-opponent space.
• Comparison of color uniformity when projected into perceptual space versus normalized cone-opponent space.

Main Results:

• Color distributions projected into cone-opponent space showed elongation along a bluish-yellowish axis, consistent with natural environments.
• Color distributions from natural environments appeared more uniform in the CIE L(*)a(*)b(*) perceptual space than in normalized cone-opponent space.
• A bluish-yellowish bias in environmental colors was identified as a potential factor shaping chromatic sensitivity.

Conclusions:

• Perceptually uniform color spaces reflect a bluish-yellowish bias characteristic of natural environments.
• This environmental color bias may be a significant factor in the evolution of chromatic sensitivity.
• Perceptually uniform color metrics can be derived from natural scene statistics and potentially adapted for specific environments.