Color constancy by category correlation.

Area of Science:

• Computer Vision
• Computational Color Science

Background:

• Color constancy, crucial for computer vision, remains an open problem.
• Existing methods often rely on physical scene constraints or statistical assumptions.
• Limited attention has been given to the impact of illuminant selection on color image representation.

Purpose of the Study:

• To propose a novel approach for color constancy using perceptual constraints.
• To leverage universal color categories derived from psychophysical measurements.
• To develop a fast method for sampling illuminants based on these perceptual constraints.

Main Methods:

• Defining a 'category hypothesis' to weight feasible illuminants.
• Utilizing universal color categories (basic linguistic terms) as perceptual anchors.
• Implementing a fast sampling algorithm for a wide range of illuminants.

Main Results:

• The proposed method demonstrates performance rivaling state-of-the-art techniques.
• The approach does not require training algorithmic parameters.
• The method proves effective in achieving stable color representations despite illuminant variations.

Conclusions:

• Perceptual constraints, specifically universal color categories, offer a promising direction for color constancy.
• The developed method provides a robust and efficient solution.
• The framework supports integration of top-down information, opening avenues for future research.