Study Of Fractal Image Compression Space Mapping

With the rapid development of computer technology, the amount resources sharing of information and the limitation of network resources, image compression/coding has become crucial technology of information communication transmission in the applications of digital communication and multimedia services. Fractal image compression coding has been attracted broad attention since it was put forward for its advantages of innovative ideas, high compression ratio, resolution independence, fast decoding, and it is recognized as one of the most promising new generation image coding technology.The matching performance of image gray value of the classic two-dimensional gray-scale transformation as only variable in the matching (similarity) calculation is poor without considering the positional relationship between the pixels in the image. And all the coefficients of the original three-dimensional gray-value transformation is performed simultaneously, then quantized and stored. In order to ensure the convergence of the decoding, the compression factor of the three-dimensional gray-value transformation will be truncated when it does not satisfy the condition (0≤s≤1) in the quantization. The truncation affects reconstruction quality and the computational complexity of encoding time is long in the original three-dimensional gray-value transformation.In response to these deficiencies, the improved space mapping gray-scale transformation method was proposed in this thesis.It put positions and brightness into the gray-scale transformation at the same time, and formatted linear mapping between three-dimensional curved surfaces. The method quantified the scaling coefficient firstly, then calculated and quantified the other coefficients of improved gray-scale transformation to improve the possibility of successful range-domain matching. The improved method is applied to Fisher quatree classification compression algorithm, experiments shows that this method reduces the number of coding blocks, then improves compression ratio and shortens encoding time while the quality of decoded image is not much influenced.According to a large number of experimental data analysis, the distribution regularities of the improvement space mapping gray-scale transformation coefficients are discovered. The thesis proposes a quantization and coding scheme which use6bit,5bit and9bit to quartzite a, b and d with comprehensive and ideal effect. Then quantization scheme was applied to the actual coding algorithm, compared with several other bit allocation scheme, the experiment result proves the rationality of the proposed scheme combining with the compression ratio and reconstruction quality consideration.Finally, the compression factors were discussed, the main analyses the effects of setting compression factors value to encoding time. Experimental results showed that reducing the s value quantization bit will decrease PSNR to a certain extent, but it can significantly reduce the encoding time and improve the coding efficiency.