||In fractal image compression an image is partitioned into ranges for each of which
a similar subimage, called domain, is selected from a pool of subimages. A typical
choice for the domain pool may consist of all square subimages of a particular size.
However, only a fraction of this large pool is actually used in the fractal code. This
subset can be characterized in two related ways: (1) It contains domains with relatively large intensity variation. (2) The collection of used domains are localized in
those image regions with a high degree of structure. Both observations lead us to improvements of fractal image compression. Firstly, we accelerate the encoding process by a priori discarding those domains from the pool which are unlikely to be chosen for the fractal code. This comes at the expense of a slight loss in compression ratio. In our empirical studies (using Fisher's adaptive quadtree method) we have found that a twofold acceleration leads to a drop of only 2 to 3% in the compression ratio while the image quality even improves by 0.1 to 0.2 dB. Secondly, the localization of the domains can be exploited for an improved encoding in effect raising the compression ratio back up without any penalty.