A study of the performance gain obtained in image coding by using eventually contractive discrete fractal transformations

In the last few years, the discrete fractal transformation has been applied to image coding in order to obtain high compression rates. The coding scheme works by dividing an image into range blocks that are approximated, through a contractive fractal transformation, by lowpass filtered versions of the domain blocks selected in the image itself. It is reported by Fisher (1994) that using eventually contractive fractal transformations one can improve the compressed image quality. However, no explanation is given to justify this fact. We address this issue by relating this improvement to the filtering process of the domain blocks.