Research On High Quality No-search Fractal Image Compression Algorithm

Fractal image compression algorithm, different from the transform coding schemes, has been extensively studied because of its high compression ratio, fast decoding and decoded image resolution-independent features, since it was proposed in the 1980s and developed in the last more than 20 years. But the time consumption is a big problem which impedes the application of this algorithm. Many search-based improve methods have been proposed but the coding speed is still very slow. Although no search fractal coding algorithms is very quickly, but the image quality is unsatisfactory.Studying on the base-line fractal image compression scheme and a large numbers of improved algorithms, the function and significance of each part of the base-line scheme is analyzed. Studies show that fewer researches have been taken on the spatial compression transform. Analyzing the distribution of the scale parameters for the range blocks with the minimum size,the phenomenon was founded that a majority of the scale parameters for the range blocks with minimum size are out of the valid range. An evaluation method is proposed by studying the nature and requirements of the spatial compression transform. An improved spatial compression method based on texture feature is proposed. In this method, the gray levels of the pixels in contracted domain blocks were increased or decreased from the mean of neighboring four pixels in corresponding domain blocks according to the difference between the value of the corresponding pixels in the range block. The proposed spatial transform method makes the range blocks easily to be matched and the collage error decreased.Substituting the traditional spatial compression method in the present no search fractal compression algorithm with the proposed method, the number of the gray scale parameter which out of the valid range is reduced significantly. Experiments show that the collage error of the novel algorithm is smaller than the present no search algorithm so the quality of reconstruction image is better because of the collage theorem, especially for the image with more high frequency components. Although the novel algorithm cost much more time encoding the image but it still much faster than the search-based algorithms.A scheme for storing the fractal coefficients is proposed and a recommended data structure of fractal compression file. With these processing, the image compression ratio is further improved. Experimental results show that the quality of reconstruction images with the novel algorithm are better than present no search fractal compression algorithms, but with respect to the search algorithms and JPEG standard algorithm there is a gap.