Biologically motivated neural computing in early vision processing

We introduce a network of networks (NoN) model to solve image processing problems in early vision. The method is motivated by the fact that natural image formation is a local process, and thus processing can be accomplished by a globally coordinated, local parallel processing structure, readily implemented by the hierarchical cluster architecture of the NoN model. The modeling is very powerful in that it achieves high quality adaptive processing, and virtually eliminates the computational difference between inhomogeneous and homogeneous conditions. Computer simulations show that this method is able to provide fast, quality image processing in early vision.