Design And Implement Of Image Compression Algorithm Based On DCT Using Hierarchical Vector Quantization

With the great development of modern science and technologies, data compression methods have been widely applied in various fields including scientific research, daily life and entertainment. As a main branch of data compression, vector quantization(VQ) is regarded as one of the effective and efficient methods adopted in theoretical analysis and applications. All its achievements can be ascribed to the three outstanding features, which are the high encoding speed, the high compression rate, and the straightforward expression framework. In image field, the method of hierarchical vector quantization(HVQ) especially keeps itself in the foreground. In this dissertation, several key problems of HVQ used in image compression are discussed. At the very beginning, we systematically summarize the goals, categorizes, theory of image compression coding as well as its current development. Briefly described the principle of one dimension and two dimensions FDCT. A well applied image compression standard--JPEG is introduced in detail, including its goals, three levels, four work patterns, basic structure and effect evaluate methods. We introduced the basic principles and quantization processes of VQ. Vector quantization can be divided into three main parts: codebook design, codeword search and codebook index assignment. The descriptions of three parts are given. The technology of HVQ adopted by this dissertation and the design process of its quantizer are described detailedly. The design and implement of HVQ used in JPEG is taken into full consideration. We put forward a new method of both HVQ and FDCT, and improved the original JPEG system. The part of mapping in original JPEG has been modified as a combination of HVQ and hierarchical FDCT. A new method of dynamic producing simple codebook and a adaptive algorithm of codeword search are applied according that pixels have relativity between each other. The former based on the technology of vector segmentation and interpolation, the latter take PSNR and MSE as evaluating indicator. The modified part has scientific process and optimized algorithm. The compression rate and the computation lessen of new system are improved obviously. The experiments of comparing modified system and standard JPEG system have conducted. A standard Lena picture, a typical picture and a batch of normal pictures are involved. The data of experiments indicated that the compression rate and the speed of computation are enhanced in the condition of keeping the reconstruct image's quality as the original. Furthermore, the modified system has a good compatibility. HVQ is now a hot research field, although has a quiet progress, but as a whole its development is not mature, there are many problems be needed more contribution. As a conclusion of this dissertation, we appoint the next research work in this field and view the future development direction of HVQ image compression.