Research On Image Compression Algorithm Based On Fractal And Wavelet Theory

Image compression algorithm has been under study,many new theories are put forward in recent years,such as sub-band coding,wavelet-transform coding and image compression algorithm based on fractal theory.The algorithm based on fractal theory has potential advantage in high compression ratio and is simple and efficient in decoding,so many researchers all around the world pay many attention on it.In addition,the standard of JPEG2000 define the standard of Region of Interested(ROI)image compression,scholars have been widely studied,too.This thesis analyzed the basic fractal image compression algorithm put forward by Jacquin and the ROI algorithm defined in JPEG2000.For the shortcomings of fractal compression algorithm which is very slow and maximum displacement method can not adjust the image quality,the following work has been done in this thesis.(1)Proposed an algorithm basis on sum-of-grad of sub-image block.Define the grad characteristic of sub-image block.Make the proof of relationship between grad characteristic and mean square error of sub-image block.With nearest neighbor in the sense of gradient search domain blocks instead of the mean square error search for a matching block,accelarating compression speed and the quality of image just has a little change.At last,the simulation results prove the effectiveness of the modified algorithm.(2)The algorithm of fractal image compression based on wavelet transform is improved.Omitted isometric transform step of spatial domain fractal compression algorithm.Dividing the wavelet coefficients into positive and negtive parts and encoding both parts separatly.This method improve the compress ratio greatly.Storing the bigger range block directly to ensure the quality of decoding image.Simulation results show that the proposed algorithm under the premise of quality assurance image decoding improves the compression ratio.(2)Make some improvement to ROI image compression algorithm.Traditional mask contains a lot of background information.In this papper,mask is determined by comparing wavelet coefficients and overcome shorcomings just talked about.The foreground of image clearer at low bit rate.At last,the simulation results prove the effectiveness of the modified algorithm.