Systems theory in modelling chromatic retinal visual processing

Abstract

Concepts and methods in non-linear System Theory, developped according to the Volterra-Wiener formulation, are one of the effective system tools to analyze certain physiological systems, and they are made apparent by setting forth formulations and solutions of various research problems of different degrees of abstraction to identify the local nonlinearity in the visual pathways.

The most outstanding characteristic of the vertebrate retina and probably of the cortex is to process data by computing layers. Based on models of layered computers to explain the coding of achromatic signals in the retina [MORE-80] we have developed a general model for the chromatic signals process.

Extending prior concepts proposed by Moreno-Díaz [MORE-79] on the mechanisms of center-periphery interaction for achromatic specialized process, we propose here a model which is essentially valid for chromatic processes by extending the input space of the ganglional computation in vertebrate retina, to permit a chromatic dimension for a retina containing receptors of different spectral sensitivity.

The model is based on the neurophysiological evidence of the channels photoreceptor-bipolar-amacrine and photoreceptor-horizontal-bipolar-amacrine, to introduce the concepts of fast and retarded signals interacting by non-linear lateral processing at inner plexiform layer and its suitable mathematical formulation. The formalization is expressed through a non-linear spatial-temporal-chromatic transformation. The results obtained are narrowband outputs ON and OFF chromatic, as well as contrast detectors in the wavelength domain: direct colour filtered signals carrying the complementary colour information. The various colour coding units appear then as consequences.

Input signals for the computer simulation correspond to the colour components RGB of real images, acquired by a camera computer system.