Color constancy

Imagine taking a white mug from daylight into a room with a warm lamp. You may still judge it to be white, although different light now reaches your eyes from it. How does vision distinguish a change in lighting from a change in the surface itself? This relative stability is color constancy, and it is not perfect in every scene.[1]

A white mug can retain a relatively stable perceived color as illumination changes, despite altered signals at the eye. How do adaptation and scene information combine to achieve this? Adaptation adjusts the sensitivity of light-sensitive cells; spatial comparisons provide information from other surfaces. A 2026 computational model inspired by the first cortical visual area, V1, learned to predict the light source from labeled images. It supports one possible operation in a model, rather than directly demonstrating that operation in living human circuits.[1] [2]

Several adjustments are known. Their complete neural combination is still incomplete.[1]

The computational result leaves open how real circuits implement illumination prediction and combine it with adaptation and other visual-processing levels. A model’s success at estimating lighting is different from a causal account of human color appearance.[1] [2]

Color helps us pick out and describe objects—a blue coat can still be the coat we want in different rooms. Stable judgments connect a changing light signal with the surfaces we use that signal to recognize.[1]

The same incoming light can admit more than one surface-and-lighting interpretation. Explaining how vision handles that ambiguity connects physical optics, experience of the scene and neural processing; it also clarifies why estimating an illuminant and reproducing human appearance judgments are different tests.[1] [2]