Research And Improvement On Rate Control Algorithm And Rate Distortion Optimization Model For H.264

The main purpose of the H.264 video coding standardization is to improve the performance of video compression. Rate distortinon optimization and rate control play an important role in video coding and communication systems by providing optimal compression performance.This dissertation has made in-depth analysis on the rate-distortion optimization scheme which has high encoding computation complexity and the foundation theory of rate control theory in H.264 video coding standard. Then, this dissertation summarized the latest research work about rate distortion model and rate control algorithm. On the basis of analysises, a new frame-level rate control algorithm was proposed in this paper. Firstly, in order to reduce the inter-dependency between RDO (Rate-distortion Optimization) and rate control, a kind of two-stage rate control algorithm was presented. Secondly, to provide more accurate prediction of the frame complexity, instead of the MAD (Mean Absolute Difference), bits information in the RDO-based mode decision process was employed to predict the frame complexity. Thirdly, a self-adaptive exponential R-Q model was proposed. The new proposed rate control algorithm overcomes the disadvantages in the rate control precision and encoding performance of former algorithms.Then an improved rate-distortion oprimization model was presented. To provide more accurate estimation of the number of coding bits for encoding intra / inter macro blocks, this dissertation also has made some work. Firstly, a new calculation method of bit rate estimation was proposed. The rate estimation is a linear combination of existing coding parameters, which are more accurately related to entropy coding and transform coefficients. Secondly, an improved rate-distortion estimation model was also presented. And then we got a new rate distortion optimization model which has superior performance.The distortion estimation is accurately related to the quantization step-size. Consequently, a new rate distortion optimization model was got which has superior performance. The dissertation selected JM14.2 as the simulation platform. The experimental results showed that the proposed RDO model can accurately estimate the bit-rate, distortion and rate-distortion cost comparing to the previous proposed models. And the proposed rate control algorithm has excellent performance both in bit rate arruracy and coding efficiency compared to the previous proposed algorithms.