Speeding up fractal image compression

Fractal coding offers many promising qualities, such as high compression ratio, good image quality, and resolution independence of the decoded image. However the encoding process suffers from the long search time of the domain block pool. Much research has been done to speed it. The fractal theory, and many speed-up techniques, such as block classification, code-book clustering, etc. are discussed in the recent book by Y. Fisher (1995). We present a different scheme to speed up the search. We devise an incremental procedure which limits the domain block pixels, given a range block and an upper bound on their distance. We arrange the domain blocks in a tree, and navigate it to select a small number of candidate blocks. Experiments show that the encoding process speeds up by up-to 50 times. Our procedure works in conjunction with other methods, such as block classification.