An inference implementation based on extended weighted finite automata [for image compression]

A similarity enrichment scheme for the application to image compression through the extension of weighted finite automata (WFA) has been recently proposed (2000) by the authors. In this paper, they first establish additional theoretical results on the extended WFA of minimum states. They then devise an effective inference algorithm and its concrete implementation through the consideration of WFA of minimum states, image approximation in least-squares, state image intensity generation via the Gauss-Seidel method, as well as the improvement of the decoding efficiency. The codec implemented in this way explicitly exemplifies the performance gain due to extended WFA under otherwise the same conditions