Similarity enrichment in image compression through weighted finite automata

Abstract

We propose and study in details a similarity enrichment scheme for the application to the image compression through the exten-sion of the weighted nite automata (WFA). We then develop a mech-anism with which rich families of legitimate similarity images can be systematically created so as to reduce the overall WFA size, leading to an eventual better WFA-based compression performance. A number of desirable properties, including WFA of minimum states, have been estab-lished for a class of packed WFA. Moreover, a codec based on a special extended WFA is implemented to exemplify explicitly the performance gain due to extended WFA under otherwise the same conditions.