Weighted finite automata, digital images and image compression

A finite resolution digital image (gray-scale) is considered to be a function with domain, the set of strings of a finite alphabet, and range, the grayness value (real) of a pixel. A finite state automaton with weights can be used to represent such a function. We have studied the algorithm, which infers the weighted FSA from the digital image, and also the algorithm for creating the digital image from the WFA. We discuss the factors affecting the compression and clarity in this representation. We describe a way to incrementally infer the WFA of an image from the WFA of another image it closely resembles.