Edge-based partition coding for fractal image compression

Abstract

This paper presents an approach for fractal image compression that yields the best performance compared to fractal methods that do not rely on hybrid transform coding. The achievement is obtained using the standard algorithm in which the image is partitioned into non-overlapping range blocks which are approximated by corresponding larger domain blocks with image intensities that are affinely similar to those of the range blocks. The particular feature of our approach is the adaptive spatial partition, that proceeds in a split/merge fashion. In previous split/merge partitions for fractal image coding the merging started from a uniform partition consisting of square blocks of equal size. In this paper we base the block merging on an adaptive quadtree partitioning. An efficient encoding method for the quadtree-based split/merge partitions is proposed which generalizes the coding of region edge maps and which includes special context modeling. The encoding method may be used for other applications that rely on image partitions.