New Technologies, Based On The H.264/avc Bit-rate Control

The third generation mobile networks have been well developed in China. Multimedia transmission is one of the most important applications over mobile networks. Due to the economical and technological constraints, the performance of video transmission is limited by the network bandwidth. There are two categories of approaches, which are used to solve the low bandwidth constraints:one is to design a brand new video compression technology and the other is to modify the existing rate control algorithms. The goal of this thesis is to explore some novel algorithms to solve these problems for H.264/AVC CODEC.H.264/AVC[9] was jointly developed by two international organizations ISO MPEG and ITU-T VCEG, which is one of the most prominent video compressed standards. H.264/AVC employs many new technologies such as multi-reference frames, high-precision motion estimation, deblocking filter, variable block sizes, multiple code modes, rate distortion optimization and so on. Meanwhile, these new technologies also produce new problems. Firstly, most of the current rate control algorithms [5][6][9][35][36][42][46] are hard to use forⅠ-frame, which is the most important component in the video streams. High qualityⅠ-frame will produce more bits than P or B frame, which will cause buffer overflow and frame skip. Low quality will greatly degraded the visual quality of the current group of pictures (GOP). Secondly, Lagrangian multiplier based rate distortion optimization is used to encode the current frame in H.264/AVC standard. In the frame level, Lagrangian multiplier is treated as a constant for each macroblock (MB) in the current frame. Although this scheme performs well in comparison with others, it is clear that const Lagrangian multiplier in frame level is not a good choice because of different characteristics for each MB. Therefore, flexible Lagrangian multiplier is required for MBs in one frame.This thesis aims to solve these two problems mentioned above. Experimental results show that the performance can be improved significantly after incorporating our proposed algorithms.