A new approach for the speed up of fractal image coding

We propose a time improved fractal image coder with a reduced domain pool and optimal luminance transform parameters calculation. This scheme is applicable to the Fischer's (1994) classification method. We accelerate the encoding process by reducing the domain pool and then by minimizing the number of operations for the similarity search; this reduction is based on discarding the domains with nearly the same variance from each class of the domain pool. This approach provides a greater speed loss with a slight loss in the compression ratio and a slight improvement in the image quality. In our experiments an acceleration of 6.7 for the image "Lena" is reached with a good decoded image quality. For a speed up factor of 2 the compression ratio is about 0.8% reduced and the image quality is about 0.23% improved. In order to increase the compression ratio again we use Jaquin's (1992) method and we remove some of the range blocks with shade property from the search.