Interframe Fast Coding Algorithm Base On Background Frame For Surveillance And Conference Video

With the rapid development of technology and network,people's demand for high-definition surveillance and conference video is increasing.The new generation of high efficiency video coding standard(HEVC)based on the H.264 standard,its compression efficiency has been increased by nearly 50%,but the coding complexity has increased significantly.In order to reduce the coding complexity of surveillance and conference video,this thesis optimizes the algorithm of the three modules of coding unit(CU)partition,prediction unit(PU)mode selection and motion estimation based on virtual background frames.This thesis proposes a fast algorithm for CU partitioning and PU mode selection based on virtual background frames.First,the pixel of original video sequence is statistically analyzed to generate virtual background frame which is used to classify the CU.Then,the partition and selection of PU mode of different types of CU are statistically analyzed.Statistical analysis shows that the partition and selection of PU mode of different types of CU have their own characteristics.Therefore,according to the partition result and texture complexity of each category of CU,a scheme for terminating the CU partition in advance is designed to skip the unnecessary CU depth traversal process.According to the characteristics of the PU mode selection of each category CU,skip PU mode detection with less possibility to reduce the complexity of the PU mode traversal process.Experimental results show that,compared with HM16.9 encoder,the algorithm reduce the coding time by 33.31% on average when BDBR increased by 1.93% and BDPSNR decreased by 0.07 d B.This thesis also proposes a fast motion estimation algorithm based on motion attribute classification of virtual background frames.First,based on the virtual background frame,the PU is divided into three categories: background PU(BPU),foreground PU(FPU),and hybrid PU(XPU),and the motion estimation results of each category PU are statistically analyzed.According to the motion estimation characteristics of each type of PU,the corresponding motion search scheme is designed.For BPUs with slow or static motion,a small diamond search method with a step size of 1 is used to narrow the motion search range.For FPUs with moderate motion complexity,a combination of hierarchical search and multi-point search is used to perform motion search in a reasonable search mode.This solution can terminate the motion search process in advance and reduce the number of search points.For XPUs with more complex motion,a combination of hierarchical search,multi-point search,and multi-predictive motion vector search is used to search to accelerate the motion estimation while ensuring the accuracy of the search.Experimental simulation results show that,compared with HM16.9 encoder,the proposed algorithm can reduce the encoding time by about 15% on average while ensuring a low loss of video quality.