Research Of Rate Control In H.264/AVC And Algorithm Improvement

In multimedia communication,compressed bits stream are usually transmitted on network of inconsistent bandwidth. In order to make sure that the compressed video bits stream is compatible with the bandwidth especially the bandwidth of low bits rate as well as the bandwidth of time-varying wireless channel and also take good advantage of the limited network resources , rate control become indispensable in video communication system. The job of rate control is to prevent the buffer from overflowing or underflowing which originated from the inconsistentness between source bits and current bandwidth.At the same time, the quality of the image should also be considered during the transmitting. Based on these reasons, the technology of rate control has always been a hot topic in the field of video coding and TM5 in MPEG-2 ,VM8 in MPEG-4 and TMN8 in H.263 are basically the classical algorithms.H.264 /AVC is the newest international video standard developed by JVT in 2003. It has gained an extensive application for its favorable coding efficiency in video communication such as handset video and high definition TV communication. However,comparing to the former video coding standard,H.264/AVC on one hand has brought in the theory of rate distortion optimization which is the root of dilemma between the"egg"and"chicken"in rate control and on the other hand,has introduced more complex coding mode which definitely increase the overhead bits such as macroblock mode bits and motion information bits. The overhead bits are various,so they will not be considered as constant any more, which is not the same as the former international standard. The character of overhead bits in H.264/AVC results in the inaccuracy of bits allocation for the frame texture and consequently the quantization parameter predicted by distortion optimization model is not so accurate as expected. Based upon these analyses,the rate control algorithm in H.264/AVC is facing a new challenge.The work of this paper mainly focuses on the improvement of overhead bits prediction and the bits allocation in frame layer. Firstly,concerning flexibility of the overhead bits,a novel scheme based chosen-probability is proposed. With an intensive exploration of spatial-temporal corelation,an optimal one between spatial correlation and temporal correlation can be chosen based upon the statistical probability. Using the corresponding type of correlation that is chosen previously, the prediction accuracy of nontexture bits can be increased. Secondly, in order to resolve the issue associated with the imbalance of frame layer bits allocation,a new complexity measure which is based upon both of the histogram of difference image and difference of histograms, is brought in to adjust the allocation of bits. when both of them are taken into account,the original rate control in H.264 could be renewed. The algorithms proposed in this paper are tested in H.264/AVC's standard test model called JM10.2 version. Experimental results demonstrate that the algorithms proposed in this paper can achieve a good performance in increasing the PSNR value and reducing the variation of PSNR, which indicates image's better quality and smoothness.