AN EFFICIENT TECHNIQUE FOR IMPLEMENTING AN IMAGE-COMPRESSION NEURAL ALGORITHM ON CONCURRENT MULTIPROCESSOR

The paper describes a pal-allel implementation of a neural algorithm p erforming vector quantization for very low bit-rate video compression on toroidal-mesh multiprocessors systems. The neural model considered is a plastic version of the Neural Gas algorithm, whose features are s uitable for implementations on toroidal mesh topologies. The architect ure adopted and the data-allocation strategy enhance the method's scal ing properties and remarkable efficiency. The parallel approach is sup ported by a theoretical analysis of the efficiency of the overall stru cture. Experimental results on a significant testbed and the fit betwe en predicted and measured values confirm the validity of the parallel approach. (C) 1998 Published by Elsevier Science Ltd All rights reserv ed.