Towards a Neural Architecture for Unified Visual Contrast and Brightness Perception

Abstract

In this contribution a brief overview is given on a computational architecture that serves as a framework for a unified theory for contrast, contour and luminance perception. The computational mechanisms utilize a center-surround antagonism based on shunting interactions which allow to multiplex local contrast as well as luminance data. This data is demultiplexed at a later stage to provide separate input to parallel but interacting processing streams for contrast and brightness information, respectively. The approach unifies the models for BCS and FCS architecture as developed by Grossberg and colleagues. Within the architecture proposed here a process for brightness reconstruction is suggested which consists of three functional stages, the activities of which are controlled by the accompanied contrast channel which in turn is build by segregation and pooling of data from the input data stream.