Biologically-Based Neural Network Model of Color Constancy and Color Contrast
Abstract
The light which reaches the eye, or any other sensor, is the product of the reflectance and the illuminant. Therefore, in order to determine the surface reflectance of an object independent of the illuminant, a system must use the spatiochromatic context of the image. We have developed a neural network based on the anatomy and physiology of the visual projection from retina to V4. The network combines color-opponent and contrast information to achieve a good degree of color constancy. This network has been tested on simulated images corresponding to the stimuli used in well established psychophysical experiments. Responses qualitatively match human responses to a variety of center-surround and Mondrian test stimuli. Color constancy is the ability to maintain an approximately constant color perception despite changes in the incident illumination of the object. Color contrast, also referred to as chromatic induction or simultaneous contrast, is the change in the (perceived) color of a surface due to the spectral composition of neighboring surface. Color perception in natural scenes depends upon both of these phenomena. Together, these two effects demonstrate that color perception does not directly depend upon the wavelength of the light reflected from a surface.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1992
- Accession Number
- ADA248128
Entities
People
- Gershon Buchsbaum
- Leif H. Finkel
- Susan M. Courtney
Organizations
- University of Pennsylvania