Segmentation and discrimination of structural and spectral information using multi-layered pulse couple neural networks

Motivated by recent biological understanding of the way in which the brain encodes discrimination information in a time signal from a large multi-layered image, temporally encoded neural networks are potentially fast and robust paradigm for fusing/segmenting multi-dimensional data. In this paper, pulse coupled neural networks (PCNNs) are applied to spectral images to determine spatial region boundaries and discriminate specific regions of interest. Several layers of pulse coupled neural networks are linked together and applied to the spectral data. Different pulse couple neural network coupling factors are applied to each network layer. The characteristic impulse times for each network is varied at each layer to optimize data segmentation. Several methods of linking the different pulse couple network layers together are also evaluated to enhance performance. The resulting impulse signals are evaluated to determine optimal parameters to enhance spectral and spatial separation of different regions