The Research And The Realization Of MPEG-4 Video Coding Based On Artificial Immune System

In recent years, multimedia has become the most important way of getting information. As one of the core multimedia technologies, video compression coding has developed fast in both theory and application for the past few years. Its progress has a deep impact on every aspects of modern society. MPEG-4 is one of the most influential international standards of multimedia information compression coding.This thesis conducts the key research on rate control and motion estimation under the MPEG-4 encoder framework.With regard to rate control, the algorithm based on quadratic rate distortion model distributes constant number of target bits among each group of pictures, and there is always inaccuracy between target bits and actual cost bits. This thesis distributes variable number of target bits which is adjusted by the difference of the target bits and actual cost bits of the previous group of pictures, so the more accurate bit rate is got. Then algorithm is realized under the framework of XVID CODEC source code.With regard to motion estimation, this thesis analyzes the algorithms based on artificial immune system which have been presented recently. Compare with the classic block matching algorithms, the algorithms based on artificial immune system improves search accuracy as a result of higher coding efficiency. This thesis designs the principle of termination, the limitation for the times of iteration and Parameters of AIS for faster searching speed. During the procedure of evolution, some antibodies produce offspring which are always restrained by each other .This does harm to save more good individuals and causes slow evolution of these antibodies. Because that,Clonal proliferation and affinity mutation are optimized with the part of DS search pattern points, so that it searches out more accurate result within limited iteration times. And this thesis uses a adaptive search window which is adjusted by motion vectors to decrease the probability that the search points go out of search range and become useless, which happens in artificial immune system algorithm usually. Overcome the weakness, the algorithm conducts better in escaping from local optimal. The improved motion estimation based on artificial immune system is realized under the framework offered by XVID CODEC. Higher PSNR of images are gained and this algorithm fits the request of real time coding.