Wavelet-based Image Compression

With the rapid development of digital communications, computer networks and multimedia technologies, image compression become a key of solution to multimedia communications and one of the most active areas in information technology. Image compression based on wavelet transform is an important branch of image compression. Against this background to study and improve the algorithms of image compression based on wavelet is no only an important task but also a research hot.The thesis is main divided into the following sections: first give a brief introduction about background, significance, study of current conditions and evaluation criterion of still image compression; and then, from the principle of wavelet transform through experiments give the characteristics of wavelet coefficients as a prepare to describe and improve the wavelet compression coding method; then introduced the wavelet image compression based on threshold. And according to the advantages and disadvantages of both hard -threshold and compression soft-threshold proposed an improved the compression scheme based on threshold. Then the research focused on the image compression based on zero-tree structure. After having a main analysis both EZW (Embeded Zerotree Wavelets Encoding for short) and SPIHT (Set Partitioning in Hierarchical Trees for short), we proposed two programs to improve SPIHT algorithm. Program I, have a pretreatment when code texture subband .Such program can improve the subjective quality of the restoration of images. Program II, Give the introduction of assumption"relationship of brothers nodes"to add the"relationship of father and son nodes"assumption; Have a different treatment to the coefficients of lowest resolution subbands and we modified the zerotree of SPIHT algorithm. Program II improve the ability of identify the significant coefficients node, so it improved the SPIHT algorithm's performance of image compression. Simulation confirmed that both program I and program II are feasible and effective.