Computational approaches to color constancy: adaptive and ontogenetic considerations

Area of Science:

• Computational vision
• Color science
• Perception

Background:

• Color constancy is essential for perceiving stable object colors under varying illumination.
• Existing computational models often simplify adaptation processes.

Purpose of the Study:

• To investigate a two-stage model of color vision for achieving illuminant-invariant reflectance estimates.
• To evaluate the necessity and contribution of adaptation at different stages.

Main Methods:

• Developed a two-stage computational model of color vision.
• Incorporated von Kries adaptation at the sensor stage.
• Analyzed adaptation at the second, reflectance channel stage.
• Used simulated daylight illuminants and natural reflectance functions.

Main Results:

• Adaptation at both sensor and reflectance stages is necessary for accurate reflectance estimation.
• von Kries adaptation partially reduces, but does not eliminate, the need for second-stage adaptation.
• Model computations demonstrate effective illuminant-invariant reflectance estimation.

Conclusions:

• A two-stage model with dual adaptation provides a robust framework for color constancy.
• The second-stage transformation plays a key role in surface reflectance estimation, comparable to opponent color transformations.