Dynamic topology networks for colour image compression

Abstract

The Self-Organizing Dynamic Graph (SODG) is a novel unsupervised neural network that overcomes some of the limitations of the Kohonen's Self-Organizing Feature Map (SOFM) by using a dynamic topology among neurons. In this paper an application of the SODG to colour image compression is studied. A Huffman coding and the Lempel-Ziv algorithm are applied to the output of the SODG to provide considerable improvements in compression rates with respect to standard competitive learning. Furthermore, this system is shown to give mean squared errors of the reconstructed images similar to those of competitive learning. Experimental results are presented to illustrate the performance of this system. © Springer-Verlag Berlin Heidelberg 2001.