Research On Rate Control Algorithm Based On SHVC Of Scalable Video Coding Standard

Rate Control has always played an important role in the video coding standards,regardless of the previous H.264/SVC(Scalable Video Coding),the current H.265/SHVC(Scalable High Efficiency Video Coding)or later H.266,rate control is always the core part of each generation of coding standards.Video coding requires rate control because the performance of the hardware device is limited,so that the signal transmission rate cannot be increased without limitation.The encoder can only transmit useful information as much as possible at a limited target bit rate,the bit rate of the transmission can be reduced as much as possible while ensuring image quality.But regardless of how the image quality is optimized,the relationship between the code rate and the distortion of the video coding must conform to the law of rate distortion,in the case where the coding rate is constant,the minimum distortion of the reconstructed video after video encoding is constant.The ratedistortion law is the theoretical basis for video coding under ideal conditions.In the case of actual coding,it is impossible to achieve,but the experimental results can be optimized.Researchers have been studying the rate control optimization algorithm for video coding,and proposed various optimization algorithms for different video coding standards.This thesis studies some problems existing in the scalable video coding standard,proposes a series of algorithms and optimizes them.The main contents are as follows:In the process of SHVC rate control coding,there is a problem that the code rate of the actual code is inconsistent with the target code rate,which is also a problem that this paper needs to deal with.Firstly,this paper establishes a hyperbolic RateDistortion(RD)model suitable for CTU-level target bit allocation by analyzing the relationship between code rate and distortion of CTU(Coding Tree Unit)image coding.And the traditional solution method is used to solve the constraint problem of the rate control model,so as to optimize the CTU level SHVC rate control algorithm.For the solution method of the rate control constraint problem,this thesis uses the recursive Taylor expansion to transform it into an unconstrained solution method,and uses the Shengjin formula to solve it.In the SHVC intra-frame coding mode,when the encoder performs rate control coding on the first frame image of the video or on the scene-converted image,the coding effect is not ideal.The reason is that the encoder does not accurately predict the code rate control coding parameters without prior information.The deep learning method has the advantage of extracting features from images.This paper will use this advantage to predict the rate control coding parameters and target bit allocation weights of the current coded image.The construction of the deep learning data set requires encoding 2,300 different images and extracting 280,000 training samples.At the same time,data fitting of the coding parameters corresponding to the sample is performed to obtain the label of the training sample.Inter-layer information is a major feature of the SHVC video coding standard.The use of inter-layer information can improve the coding efficiency of the encoder for the enhancement layer image.However,in the rate control process of the SHVC video coding standard,there is a problem that the coding information of the base layer is not applied to the enhancement layer coding.In this paper,the distortion obtained by the base layer coding is extracted,which is used as the inter-layer information to adjust the target bit allocation process of the enhancement layer.At the same time,a depth learning based rate control algorithm is applied to the base layer.The comparison of the experiments in this thesis is based on the use of the inter-layer information method,and compares the rate control optimization algorithm of whether to join the deep learning method.