Parallel answer-in-weights neural network implementation for image preprocessing
Iren Valova, Y. Kosugi · 2002
A neural network for image preprocessing is utilized for a parallel neurocomputer. The algorithm is based on an inverse approach and provides lossless decomposition and high rate compression of grey scale pictures with dimensions 128/spl times/128 pixel. The basic properties of the method are discussed and demonstrated by examples. The network is easily expandable for the purposes of image segmentation or feature extraction.