Research On Algorithm Of HEVC Loop Deblocking Filter

The new generation of High Efficiency Video Coding(HEVC)standard has a higher compression rate than the previous generation video coding standard.However,excellent coding performance is usually accompanied by a huge coding complexity.Moreover,block-based coding can cause blocking artifacts which seriously affects the subjective and objective quality of the video and the user experience.Therefore,loop deblocking filter is used to reduce the blocking artifacts in the HEVC standard.But the complex deblocking filtering process is not conducive to real-time video applications.In this thesis,two kinds of HEVC adaptive loop deblocking fast algorithms are proposed for the redundancy calculations and steps in the deblocking filter of HEVC loops.Through statistics and analysis,it can be seen that the filter conditions of the two boundaries within the same filter unit have a strong consistency.Moreover,the clipping function used to limit the pixel correction amplitude in the filter process greatly increases the amount of calculation.Therefore,this thesis proposes a HEVC deblocking filter algorithm that combines the basic filter unit calculations.The basic filter units are combined firstly.Then calculate the pixel change rate and other parameters by way of cumulative sum of absolute difference.Then make filter decisions on the boundary.Finally,the pixels are corrected by using a strong filter algorithm of infinite amplitude function or an original weak filter algorithm.Experiment results show that the algorithm reduces the deblocking filter time by 18.001% and the decoding time by 6.875% on average.And it has little effect on the subjective and objective quality of the video.The purpose of improving the deblocking filter performance is achieved.Obtaining the boundary strength is a key step in deblocking filter.However,redundant calculations are performed on the boundary of the same situation.Moreover,there are much cases in which the pixel values before and after filter remain unchanged or the quality of the filtered video is worse.There are only two types of filter,strong and weak,which are less adaptive.Therefore,this thesis proposes another HEVC deblocking fast algorithm based on pixel absolute difference classification.First,the number of calculations of the boundary strength is determined by the division mode and the depth of the neighboring blocks.Then,the filter boundary is filtered with the absolute difference of pixels at the boundary and the threshold.Next use pixel absolute difference to obtain the pixel change rate of the filter block.The filter boundary is divided into four modes according to the adaptive threshold.Finally,different filters are used to correct the pixels for different modes.Experiment results show that the algorithm can not only keep the subjective and objective quality of the video unchanged basically,but also can reduce the deblocking filter time by 29.394% and the decoding time by 11.901% on average.The algorithm effectively reduce the certain filter and decoding time.