Local iterative improvement of fractal image codes

In fractal image compression, the code is given by a contractive affine mapping whose fixed point is an approximation to the original image. Usually, the mapping is found by the collage coding method. We propose an algorithm that starts from an initial mapping obtained by collage coding and iteratively provides a sequence of contractive mappings whose fixed points monotonically approach the original image. Experimental results show that the rate-distortion improvement over collage coding is significant.