A theoretical framework for retinal computations: insights from textbook knowledge

Abstract

Neural circuits in the retina divide the incoming visual scene into more than a dozen distinct representations that are sent on to central brain areas, such as the lateral geniculate nucleus and the superior colliculus. The retina can be viewed as a parallel image processor made of a multitude of small computational devices. Neural circuits of the retina are constituted by various cell types that separate the incoming visual information in different channels. Visual information is processed by retinal neural circuits and several computations are performed extracting distinct features from the visual scene. The aim of this article is to understand the computational basis involved in processing visual information which finally leads to several feature detectors. Therefore, the elements that form the basis of retinal computations will be explored by explaining how oscillators can lead to a final output with computational meaning. Linear versus nonlinear systems will be presented and the retina will be placed in the context of a nonlinear system. Finally, simulations will be presented exploring the concept of the retina as a nonlinear system which can perform understandable computations converting a known input into a predictable output.