HEVC Fast Adaptive Sample Offset Technology Based On Time Domain And Spatial Correlation

With the increasing demand for high-resolution video,the development of high-compression video compression algorithms has become more and more urgent.Because of its excellent compression efficiency and compression quality,HEVC(High Efficiency Video Coding)video compression algorithm has become the current mainstream video coding standard.HEVC uses a coding structure similar to H.264/AVC(Advanced Video Coding)and introduces several new coding techniques.Among them,SAO technology(Sample Adaptive Offset)is located after deblocking filter and is used to eliminate the ringing effect caused by the quantization.Through the study of the ringing effects characteristics,it can be known that the root cause is the loss of high-frequency components of the video signal due to quantization,and the expression pattern is mostly a texture phenomenon in areas where the image brightness or chrominance gradient values are large.In this paper,proposed algorithm based on the HEVC video compression coding standard,the SAO algorithm is the main research object.Starting from the intra prediction direction and the time-domain similarity of the image,the standard SAO algorithm is modified accordingly to Pre-selection of SAO mode.At the same time,this paper proposes a fast SAO mode selection algorithm,which can greatly reduce the SAO encoding time without affecting the SAO compensation effect.The main work of this article can be summarized as the following three points:1.A fast selection method of EO mode based on intra-prediction direction is proposed.This method combines the direction of the intra-prediction angle mode to pre-process the SAO.According to the intra prediction mode of each PU,a histogram of each PU(Prediction Unit)in the currently-to-be-compensated CTU is used to obtain a CTU intra-frame prediction direction and is mapped to a corresponding EO mode.This method can reduce 79.5% SAO encoding time in low-delay P frame coding mode.2.The EO(Edge Offset)mode fast selection method based on time domain and airspace similarity is proposed.This method combines the similarity between adjacent frame and intra prediction mode and EO direction to preprocess the SAO algorithm.The algorithm obtains a Coding Tree Unit(CTU)having the highest similarity degree of adjacent frames based on the MV,and processes the currently-to-be-compensated block in the CTU compensation mode.Exclude orthogonal EO modes.The above two methods can reduce the average encoding time by 19.8% and 18% respectively.3.A fast SAO mode selection algorithm is proposed.This algorithm integrates two kinds of EO mode selection algorithms: time domain and airspace,further optimizes the SAO algorithm process.This algorithm can reduce the SAO encoding time by an average of 36.7% on the premise that BD-rate increases by 0.0326% on average.Taking into account the temporal and spatial correlation of image sequences,this paper uses inter-predictive MV and intra prediction modes to preprocess SAO,and proposes a fast SAO mode selection algorithm.Compared with the compression model provided by HEVC,the proposed algorithm greatly reduces the SAO encoding time without affecting the SAO compensation performance.