Image Compression Based On Multiscale Analysis

The image information plays an important role in modern communication. We need to compress the image which has huge data in order to save the storage and improve the efficiency of the channel. In 1980s, profiting from the wavelet theory development, the image coding has a better performance. The efficiency of image compression can be improved very much by using wavelet transform. In addition, this multiresolution transform is free from block effect in case of the Discrete Cosine Transform (DCT). With an embedded bit stream, the reception of code bits can be stopped at any point and the image can be decompressed and reconstructed.A new modified embedded zerotree wavelets incoding(EZW) algorithm is proposed after reseaching several popular embedded coding algrithm. The experiment results show that the new algorithm improved the peak signal noise ratio (PSNR) of the reconstructed image. A simple construction of second generation wavelets, the lifting scheme, is presented. Such lifting scheme can lead to a faster, in-place calculation of the wavelet transform. And the experiments testify the lifting schem is better than traditional wavelet.wavelets are good at catching point singularities, but fail to represent efficiently singularities along lines or curves. The ridgelet transform is the best approximation method for the image with linear singularity. The idea is to map a line singulariy into a point singularity using the Radon transform. The relativity between the ridgelet coefficients is discussed. And an algorithm for image compression based on ridgelet transform is designed. The experiments results show that the proposed approach can achieve higher compression ratio and PSNR, as well as better visual result.Ridgelet representation solve the problem of sparse approximation of smooth objects with straight edges. However, in image coding, edges are typically curved rather than straight and ridgelets cannot yield efficient representation. A new image representation, Contourlet transform, is proposed. The Contourlet transform, which is multiresolution and multidirectional expansion, can be applied efficiently to capture smooth contours. A comparative study between the contourlet and the wavelet analysis and the potential of the contourlet transform for image compression are discussed.