Redundancy Elimination Among Global Similar Blocks And Image Set Coding

With the development of information technology,the importance of image compression has become increasingly prominent.At present,the mainstream image compression method mainly focus on reducing the redundancy within the image.If the correlation between images can be fully utilized,the space occupied by the image can be further reduced and a better compression effect can be achieved in the case of storing or transmitting large number of images.This thesis first expounds the background and significance of the research and the relevant theories and technologies.Based on the existing research,theory and technology,this thesis proposes a universal image set compression algorithm,which can achieve better versatility,based on similar block redundancy elimination in global area.By cutting images into blocks of various sizes at different levels,the algorithm makes full use of the respective advantages of coarse grain and fine grain,finding and replacing similar blocks globally.The record of replacement is outputted by the replace procedure.The classical discrete cosine transform is used to compress the remaining blocks that are not replaced to eliminate the redundancy more sufficiently.For the generated replacement records,run-length,Huffman and exponential Columbus methods are used to encode.The coefficients generated after lossy compression of the residual block are divided into DC and AC parts,and the DC part is encoded by DPCM method,and the AC part is encoded by the run-length coding and Huffman coding after quantization.A corresponding decoding algorithm is also proposed for the encoded file.To reduce the time spent by the procedure of replacement,a hash-based algorithm is proposed.The idea proposed by this thesis,which eliminates redundancy by the way of block cutting,can treat images with different resolutions and has strong universality because training in advance is not needed.Experiments show that in the condition of same bit rate,the image quality of the proposed algorithm is better than JPEG and the algorithm based on sparse representation.In addition,the hash algorithm proposed by this thesis can effectively reduce the time of block replacement without reducing the replacement rate too much.