Evolutionary Cellular Automata for Image Compression

Abstract

A technique for image compression using binary Evolutionary Cellular Automata (ECA) is presented. The method is applied to black and white fingerprint images. These ECA are evolved with a Genetic Algorithm (GA) guided by a fitness function consisting of a similarity measure between the configurations of the ECA and the target image. When a suitable approximation of the target image is achieved it can be codified with the rule numbers and logical operations linking the ECA. In this way the original image undergoes a compression process since it will be represented by a considerably smaller bit amount. Typically compression rates of the order of 100:1 are achieved in the experiments. A codified or compressed image can be regenerated recovering a very good approximation with a typical similarity of 96% to the original image. The method is compared in speed and compression rate with other commercially and widely used compression algorithms.