The Research Of Algorithms For Block-Based Image Coding And Codebook Design

In this paper, we study the algorithms for zerotree wavelet-based image coding and codebook design for vector quantization thoroughly. A improved low bit rate image compression algorithm, including an united coding and a block pretreatment, is put forward about still images based on embedded wavelet zerotree image coding scheme. The reconstructed images are good, and the peak signal-to-noise ratio of the reconstructed images is improved compared to the original algorithm at the same bit-ratio. In this paper, we also make the classical algorithms for vector quantization improved, and improve the convergence rate and the performance for codebook design greatly. Through the experimental and simulated results, we can demonstrate that the maneuverability and the practice value of the algorithms put forward in this paper are good.Three parts are included in the thesis:1 Selection Method of the Optimization Wavelet Base in Image Compression Based On Wavelet Transform. Different wavelet base has different time-frequency characteristics. There is a question of selection method of optimization wavelet base in the application of wavelet analysis. This paper reveals several important features about the selection of the wavelet base. It tells that we should grasp the features of wavelet base in its application through actual examples. It shows how to choose the suitable wavelet base according to the needs of application.2 A new algorithm based on the Adaptive United Blocked Embedded Zerotree Wavelet(called as AUBEZW). In the Embeded Zerotree wavelet Algorithm, there is a restrict of the relativity between "parents" and "children". As the relativity become less and as the threshold falls, the possibility of the occurrence of the zerotree will fall. And the sum of zerotree root decreases, while the sum of the isolated zeros increases. So the bit sum of output data stream increases. In this paper, we propose the AUBEZW Algorithm, as the improved algorithm of the EZW Algorithm. We have make these works: When the image is processed by wavelet transform, we make block pretreatments aimed at the high frequency band of the image. After that, we discuss a coefficient type optimization design. The following coefficient coding and swing coding processes are based on the blocking coding process. Then, we propose an united coding of nearby coefficient symbol and importance. At last, we make a Adaptive Arithmetic Coding process. The new method make a mass of nonimportant coefficients centralized in image blocks, which can use the waveletcoefficients more effectively. Through the optimization process and the Arithmetic Coding process, we can decrease the code rate efficiently and enhance the visual effect of the reconstructed image(subjective and impersonal).3 A New Efficient Algorithm for Codebook Design. LBG algorithm is a classical algorithm for vector quantization. In this paper, we make it improved, and improve the performance of entire algorithm combined with improved PCC algorithm greatly. The results show the entire algorithm has fast convergence rate and excellent performance for codebook design. The improvement of PSNR for the proposed algorithm is great compared with that of the LBG algorithm.