Neural networks for image coding: a review

Images are forming an increasingly large part of modern communications, bringing the need for efficient and effective compression. This survey aims to give the reader a feel for the wide range of ways in which neural network technology has been applied to image compression. Many variants of vector quantisation are presented which use the topology preservation of the self-organising map in a variety of ways. Image coding using neural networks to guide a classical coding technique have been largely overlooked. Some promising contributions in this area are also discussed. Finally, a range of transform and predictive coding techniques are described.