A unified neural framework for early visual information processing

This paper proposes a unified neural network framework for the modeling and process of early visual information. The proposed framework is based on a neural network with hierarchical cluster architecture (NNHCA) which nicely simulates the regionally coordinated locally parallel processing structure embedded in the image formation process. The internal structure of NNHCA is quite flexible so that a large number of processing techniques can be accommodated within the framework. Examples on restoration and enhancement are provided to demonstrate the performance of the proposed method.<>