Local Adaptive Learning Algorithms For Blind Separation Of Natural Images

In this paper a neural network approach for reconstruction of natural highly correlated images from linear (additive) mixture of them is proposed. A multi{layer architecture with local on{line learning rules is developed to solve the problem of blind separation of sources. The main motivation for using a multi{layer network instead of a single{layer one is to improve the performance and robustness of separation, while applying a very simple local learning rule, which is biologically plausible. Moreover such architecture with on{chip learning is relatively easy im-plementable using VLSI electronic circuits. Furthermore it enables the extraction of source signals sequentially one after the other, starting from the strongest signal and nishing with the weakest one. The experimental part focuses on separating highly correlated human faces from mixture of them, with additive noise and under unknown number of sources.