Region evolution with non-linear block transformations for fractal image coding

This paper presents an investigation carried out on fractal or attractor image compression covering the aspects of lower bit rates and better image quality. Non-linear block transformations exploiting the second order polynomial are employed to improve the approximation ability over linear models for domain-to-range matching. Though better reconstructed image fidelity has been achieved, extra code arises due to the increasing coefficients. To combat this, block-based transformations are extended and pruned to region-based ones in order to reduce the number of transformations needed. Consequently, it offsets the aforementioned overhead and obtains higher compression ratios. Experimental results suggest that the proposed approach is favorably comparable with other fractal based compression methods