What image features guide lightness perception?

Area of Science:

• Visual Perception
• Computational Neuroscience
• Psychophysics

Background:

• Lightness constancy, the perception of surface color under varying illumination, remains poorly understood.
• Existing theories propose different image features crucial for lightness perception, leading to significant debate.

Purpose of the Study:

• To investigate which image regions influence human lightness judgments.
• To compare human lightness perception with predictions from four computational models.
• To test and refine existing models of lightness perception.

Main Methods:

• Classification images were measured for human observers and four computational models (high-pass filter, oriented difference-of-Gaussians, anchoring, atmospheric-link-function).
• Observers judged the relative lightness of test patches within variants of the argyle illusion.
• Model performance was evaluated based on their ability to predict human judgments and their generated classification images.

Main Results:

• Human lightness judgments were influenced by local, anisotropic stimulus regions adjacent to uniform lighting areas.
• The anchoring and atmospheric-link-function models successfully predicted human lightness illusions.
• All four models produced classification images distinct from human observers, indicating different guiding image regions.

Conclusions:

• Human lightness computations can be both local and dependent on lighting boundaries.
• The study provides a novel method for testing and differentiating models of lightness perception.
• Findings suggest that mid-level visual processing plays a critical role in lightness constancy.