Improvement Algorithms For Rate Control Based On H.264/AVC

Any video coding system can do without rate control, this is just because: ratecontrol not only control the output stream of the video coding system according tothe constraints of the signal channel to guarantee the video information transmissionquality, but also ensure that the decoder can obtain an optimal video picture quality.H.264, as the most popular video encoding standard at present, adopts many newcoding technologies compared to the previous standards, such as model selectiontechnique based on rate distortion optimization. These new technologies improve itscoding efficiency, meanwhile bring some difficulty problems to the process of ratecontrol of H.264. Therefore, lots of the rate control algorithms of other videocompression encoding standard are not applicable in H.264. Although someclassical rate control algorithms of H.264have done these problems, they still havesome other problems, such as the QP selection of I frame, target bits allocation, theprecision of MAD prediction and the high algorithm complexity.In this dissertation, for solving these problems, the H.264encoding technique,the classic rate distortion model and the JVT-W042rate control algorithm areanalyzed. The improvements for H.264rate control are done based on this analysis,as shown below:First, a new rate distortion model is derived by using the rate distortion theoryand with the help of image quality assessment method. Then according to theproposed model, some improvements for the QP selection of I frame rate controland for the target bits allocation and encoding complexity of P frame rate controlare done for the existing problem of JVT-W042in the mode0, and then somesimulations are done on the JM16.0model. The results of simulations show that thenew algorithm not only causes the target bits allocation of I frame and P frame, theQP selection of I frame to be more suitable, but also omits the MAD linearprediction process by using one new coding complexity instead of MAD, andeliminates the step of mapping the quantization step to the quantization parameterQP, and reduces the parameters of rate distortion model, and thus it reduces thecomplexity of the algorithm as well as obtains better rate control effects.For precisely controlling the status of the buffer and the target bit rate, PIDalgorithm is adopted in the H.264rate control based on the proposed model. Theerror between the actual buffer fullness and the target buffer fullness is the controlvariable in this rate control algorithm based on PID. Then the target bits allocationin the frame layer rate control are corrected and the control aim of QP is achieved.The simulation results of this method on the JM16.0model show that this method makes the actual buffer fullness fluctuate in the target buffer, and improves thePSNR, meanwhile, reduces undesirable skipping frame number and PSNRfluctuation, and then achieves better rate control.