Adaptive block partitioning in fractal image coding

We show a practical fractal image compression method using adaptive block partitioning to achieve the optimal block partition. The method is the same as the one proposed by Barnsley (1993) except for the way in which the image is partitioned. In our method, firsts we evaluate the expected maximum local collage error for each domain block, before the image compression processes. Then we both locally refine one range block to four equal sizes smaller blocks, and rejoin four equal size blocks to one larger block, recursively. We do this until all the local collage error of the domain blocks become close to the average local collage error in the image. We call this the equi-distribution algorithm. We fix the number of the domain blocks in both the refining and rejoining domain block at the same time. We repeat this process to get the minimum collage error of the image. Based on our method, it is possible to obtain the optimal domain partition for the desired file size