Research Of Rate Control Models And Algorithm

The increasing requirement of video communication encourages the development of video compressing techniques.To obtain the best visual quality under the limited bandwidth and produce an output bit stream matching the desired bit rate and buffer size, rate control is employed in video compression technology. Since H.264/AVC standard employs Rate Distortion Optimization (RDO), it makes rate control in H.264/AVC more complicated than those in other standards such as MPEG-2 and MPEG-4.It induces hot research issues in H.264/AVC rate control how to precisely control visual buffer fullness, reduce the computing complexity and obtain better visual quality.This thesis systematically analyzes the basic theory of video compressing standards and discusses three important rate control algorithms in H.264/AVC. Based on the conclusion and reference of classic rate control algorithms, we propose some novel models. Furthermore, an incremental rate control algorithm is proposed. Extensive experiments demonstrate that the new rate control algorithm can efficiently control the virtual buffer, precisely achieves target bitrate, obviously reduce unexpected frame-skip, improve and smooth visual quality. Our researches include the following parts:1.Based on the traditional R-D models, this thesis analyzes the relationship among bitrate, picture coding complexity and quantization parameter. According to the theoretical derivation and the experimental simulation, a new exponential R-D model is proposed. The experimental results demonstrate that the new model precisely describes the R-D relationship in H.264/AVC.Since the exponential model has invariant derivation, it performs well in the computation of incremental quantization parameter.2. We propose a new computation method for the picture coding complexity of intra-frames and a prediction approach for the picture coding complexity of inter-frames. The image gradient and the statistics of the histograms are employed to calculate the picture coding complexity of intra-frames.Based on the proposed exponential R-D model, this thesis makes use of the harmonic mean of the picture coding complexity for coded frames to predict the picture coding complexity of the current inter-frame. The simulation results indicate that these two approaches are efficient and accurate. 3.Based on the analysis of the relationship between distortion and quantization parameter, we propose a new exponential PSNR-QP model.The characteristics of the novel model are discussed. It provides the theoretical basis to obtain the smooth visual quality for B-frames.4.Based on the exponential R-D model and the novel prediction method of picture complexity, we propose an improved incremental rate control algorithm, which includes the weight distribution for the bit allocation of intra-frames and a proportional-integral-derivative (PID) controller for the bit allocation of inter-frames. The rate control algorithm is composed of QP calculation in GOP level, frame level and MB level. It also includes the QP calculation for B-frames which is used to smooth the visual quality.5.We implement the proposed incremental rate control algorithm in the reference software JM13.2 of H.264/AVC.Equivalent test conditions are configured for both JVT-W042 and the proposed rate control scheme, and we test extensive video sequences in various coding structures. The simulation results demonstrate that our scheme outperforms JVT-W042.The overall performance of the proposed algorithm is excellent. And it achieves better visual quality.