Handbook for Digital Signal Processing

third part is the most interesting one, containing thoroughly described application examples and general analysis on architectural aspects. The fourth part contains two neural network papers. Paper 15 presents a parallel approach for texture generation using random neural networks. Each neuron interacts with the neighbouring neurons to produce its local contribution for the complete regular texture. Paper 16 presents a two-layer system for human face recognition. The processing layer consists of three parallel neural networks, each optimized for a different recognition task. The outputs of these subnetworks are then weighted by a supervisor network to obtain the operator output. The emphasis of Parallel Image Processing is strongly on the image grammars for pattern representation and recognition. I personally would have liked to see more papers dealing with the actual implementations, especially using parallel computers or processor farms. Now the book can be mainly recommended for persons working on research related to the grammatic representation of images. The papers selected for this publication are without questions of high quality and originality but a certain lack of comprehensiveness in the first two parts makes some of the papers difficult to read.