Optimization For New Generation Video Coding

With the development of video application,high efficiency video coding(HEVC)standard cannot satisfy the practical requirements any more.Therefore,a collaborative joint video exploration group composed by the ITU-T video coding experts and the ISO/IEC moving picture experts group,dedicated to developing new coding standards beyond HEVC,and finally formulated a new standard,i.e.,versatile video coding(VVC).VVC uses the same hybrid video coding framework as HEVC,which consists of many coding modules,such as rate control,prediction,transform coding,entropy coding,loop filtering,and so on.In this thesis,we optimize the partition and rate control of VVC so that we can effectively improve its performance.To achieve our goal,fast block partition algorithm is designed to reduce the encoding complexity and the key-frame difference is utilized to improve the rate-distortion(R-D)performance of the rate control.The details are as follows:1.Compared with HEVC,VVC achieves better rate-distortion performance and the bitrate can be decreased by more than 30% at the same quality.However,the new techniques using in VVC dramatically increases computational complexity.In order to make VVC more suitable for practical applications,it is very necessary to reduce its complexity.Block partition is one of the most time-consuming procedure in VVC,optimizing it can save lots of encoding time.Firstly,we study the relationship between video content and block partition.Then,we use the gradient difference and content difference of subblocks to guide the decision of VVC intra-frame block partition so that we can skip some unnecessary partition methods,and significantly reduces the encoding complexity.2.In order to improve the performance of rate control,we study the influence of key-frame quality on its subsequent encoded frames,and proposes a rate control algorithm based on the key-frame differences.This method explores the relationship between the key-frame difference and the bit allocation factor,and it establishes the optimal bit allocation factor determination model.During the encoding process,the model adaptively determines the bit allocation factor of key-frames,which improves the R-D performance.We offline train our model,which effectively reduces the encoding complexity.Meanwhile,the adaptive determination of bit allocation factor guarantees a high allocation accuracy.