Research On Rate Control Techniques For H.264Video Coding

As the most popular video compression coding standard, H.264is one of the research hotspotsof the industry for its high compression ratio, high image quality, strong fault tolerance and strongnetwork adaptability. Rate control is the core part of the video encoding technology. By adjustingthe encoding parameters, it can make the output encoded stream adapt to the channel bandwidth andstorage limits, and at the same time achieve the maximum image quality. Therefore, the research onrate control technology is of great significance.Based on in-depth analysis of H.264basic unit rate control algorithm, the thesis improves thealgorithm from the following three aspects:(1) To address the inaccuracy of basic unit MAD prediction, an improved algorithm is proposedthat the predicted MAD is calculated by a combination of linear prediction model and thespace-time weighted prediction model, and corrected according to the prediction error of theprevious basic unit.(2) To address the inaccuracy of basic unit complexity estimation, it is proposed to estimatebasic unit complexity by a combined PSNR-based and MAD-based method.(3) To address the high computation and inaccuracy of bit allocation caused by adopting a largenumber of prediction information, a novel bit allocation algorithm is proposed based on the encodedinformation.The simulation results show that the improved MAD prediction model achieves more preciseprediction accuracy and better rate control performance than the linear prediction model. Comparedwith JM8.6, the improved basic unit layer rate control algorithm not only obtains more accurate ratecontrol but also obtains higher PSNR. At the same time, the PSNR fluctuation and encoded outputbits fluctuation are gentler, and the buffer fullness is closer to the target value.