Several Approaches Of Improved Fast Fractal Image Compression

At present, wavelet technique and fractal theory are becoming two hotspots in the field of image compression. Although JPEG2000 originating from JPEG, which is short for Joint Photographic Experts Group, has been regarded as the international standard for image compression, the low compression ratio can’t satisfy the need of image transmission and storage in the process of big data. Therefore, fractal image coding (FIC) with high compression ratio and resolution independence is studying widely. Because of the huge time-consuming, more improved FIC methods are proposed, which are at the expense of compression ratio for encoding time and reconstruction quality. It comes to the conclusion that if FIC method mixes other acceleration approaches, such as evolutionary algorithm, wavelet technology and spatial correlation, it can reduce the coding time largely and improve coding efficiency. Base on this conjecture, the paper proposes three accelerated hybrid FIC algorithms:(1) An efficient FIC method based on discrete wavelet transfer and simple range classification is proposed. Before encoding, images with one-level wavelet transform are first partitioned into four subbands, which classify the range blocks as four types. Low frequency components employ a method called variance sorting scheme with high reconstructed quality. Hereafter, the low frequency part rebuilt performs one-level wavelet decomposition once again. The first level high frequency subbands excluding the diagonal directions are predicted according to the next coarser frequency scale in the same directions. Experimental results compared with other classification algorithms demonstrate that the proposed method can obtain high compression ratio and reduce the encoding time without significant loss in the reconstructed image quality.(2) A novel FIC algorithm using particle swarm optimization and hybrid quadtree partition scheme is introduced. A method called particle swarm optimization strategy based on range block classification is presented instead of utilizing the method in the whole range pool. This new idea can enhance the compression ratio significantly and speed up the encoder. Moreover, in order to improve the quality of the retrieved image further, combining quadtree partition scheme is adopted. The Simulation results show that the proposed algorithm can obtain good quality and higher compression ratio.(3) A novel FIC algorithm based on domain blocks sorting strategies and modified no search scheme is presented. Firstly, the origin image is divided into sub-images using adaptive quadtree partition scheme. Secondly, one finds the location of the best matching domain block for the range blocks, whose sizes are larger than the preset minimum value and adopt modified no search strategy. Thirdly, the types of the range block and domain block are computed employing the proposed approach and then the corresponding computation of mean square error is determined. The computation of the mean square error is reduced and the encoding phase speeds up. Moreover, the range blocks with the minimal sizes are encoded applying the proposed domain blocks sorting method. Contrast experiment results show that the proposed algorithm can obtain good quality of the reconstructed images and shorten the encoding time significantly.In conclusion, the three algorithms above with different features improve the traditional FIC methods from different views. Moreover, the three algorithms with distinctive characteristic have their own application territories. The first method uses wavelet technology, which overcomes the block effects and speeds up the encoding phrase compared with other accelerated schemes, such as FIC method based on DCT. The method is appropriate for the occasion of high-quality. The second method adopts PSO strategy, which reduces the coding time and enhances the compression ratio compared with other FIC methods. The method is suitable for the situation for high compression rate. The third method utilizes the modified no search scheme, which can obtain high retrieved images in comparison with other FIC algorithm using conventional no search scheme. The method is appropriate for the occasion for the high-quality and rapid encoding.