Anthropocentric
Realism in Three-Dimensional Scenes
Psychology & Neural Science, New York University
laurence.maloney@nyu.edu
Biological
color perception is often compared to ideal color algorithms
that attempt to estimate the surface properties of objects
corresponding to the subjective impression of color (“intrinsic
colors”) despite changes in scene illumination and content.
Recent work in philosophy by Hilbert and others presupposes that
biological visual systems can accurately estimate intrinsic colors at
least under some conditions. Such
algorithms can only work to the extent that there are physical
constraints on possible surfaces and lighting in scenes that permit
recovery of these “intrinsic colors”.
I’ll briefly review
earlier work that focused on perception of flat Mondrian scenes and then
review more recent work that considers the problems posed to an ideal
visual system in three-dimensional scenes with non-uniform illumination
and report a series of experiments evaluating how well human observers
estimate color and lightness in such scenes and how well human observers
estimate surface properties analogous to color such as gloss and
roughness. I will also discuss a recent experiment that challenges
whether the subjective impression of color corresponds to an imperfect
estimate of surface properties. The results of this experiment
indicate that perceived color is a systematic representation of surface
properties and also illumination conditions.
|