Lean domain pools for fractal image compression

In fractal image compression an image is partitioned into ranges for each o f which a similar subimage, called domain, is selected from a pool of subim ages. However, only a fraction of this large pool is actually used in the f ractal code. This subset can be characterized in two related ways: (1) It c ontains domains with relatively large intensity variation. (2) The collecti on of used domains is localized in those image regions that show a high deg ree of structure. Both observations lead to improvements of fractal image c ompression. First, we accelerate he encoding process by a priori discarding he low variance domains from the pool that are unlikely to be chosen for t he fractal code. Second, he localization of the domains may be exploited fo r an improved encoding of he domain indices, in effect raising the compress ion ratio. When considering the performance of a variable rate fractal quad tree encoder we found hat a speedup by a factor of 2-3 does not degrade the rate-distortion curve ranging from compression ratio 5 up to 30. (C) 1999 SPIE and IS&T. [S1017-9909(99)00301-3].