Diffusion of the attractor of fractal coding for edge restoration

Diffusion of the attractor or reconstructed image of the fractal code provides a technique to restore edge information. Because of coding error associated with the fractal mappings, edges are degraded at high compression ratios. Partitioning compensates for the degradation, but lowers the compression ratios significantly and does not insure the retention of significant edges. The diffusion technique uses the image gradient to control the rate and direction of diffusion. This allows for smoothing in flat (low intensity transitions) regions and sharpening in edge (high intensity transitions) regions. The usage of the image gradient in this method insures the retention of significant edges. The diffusion technique presented in this paper lessens the degree of degradation of edges from fractal coding at a lower bit rate cost than partitioning at small block sizes.