Abstract

Spatially organized center-surround spectrally opponent retinal receptive fields serve the purpose of simultaneously coding both spatial and spectral visual information. The spatial center-surround structure serves to reduce spatial signal redundancy¹ and the spectral opponent response serves as a mechanism to reduce correlation between receptor outputs.² The spatial and spectral responses were separately deduced from the hypothesis that the purpose of retinal signal coding is redundancy reduction and compression of spatial and spectral image information. However, the spatio-spectral response of retinal receptive fields cannot be written as a product of separate spatial and spectral responses. Such separation would result if spatial and spectral information were independent. This property of receptive fields is related to properties of natural images. While the actual color of an object could be considered independent of its spatial properties, variations in spatial details would be associated with variations in spectral details, because different spectral projections of an image are spatially correlated. It is argued that the spatio-spectral structure of receptive fields can be inferred from the hypothesis that their purpose is to simultaneously and efficiently code both spatial and spectral information in natural images.

© 1985 Optical Society of America