Research On In-loop Deblocking Filter For High Efficiency Video Coding

With the high-speed development of video information, people are increasinglystrict with the requirement of high definition, high speed video quality. Currentdomestic and international coding standards are block-based, what coarsequantizations following the blocking processing seriously affect people’s vision isblocking artifacts. This paper is based on a new generation of video coding standard–the High Efficient Video Coding (HEVC). With regard to blocking artifacts affectingthe video quality, in-loop deblocking filter technologies for HEVC is studied, startingfrom the innovations of HEVC that compared with the standards before.In this paper, the in-loop deblocking filter algorithm principle for HEVC isfirstly elaborated, using HM software, compare and simulate the video sequences as asolid foundation for subsequent tests. Basic experiments show that for certainsequences, especially high complexity video, usually a very satisfactory result can notbe got. Therefore, this paper proposes two adaprive in-loop deblocking filter whichfrom the perspective of enhancing adaptability.Firstly, a content-adaptive deblocking method is proposed. By edge informationobtained through the global orientation energy edge detection and partition, filterthresholds, appropriately to best preserve image details while reducing blockingartifacts are designed. This method is characterized by (i) Maintain DCT compressionperformance for video while improving the visual quality.(ii)Compare with theluminance component, and even slightly better on the chroma components.(iii)Ourmethod especially benefits sequences with directional edge regions and large smoothregions, which are quite common in high-definition video contents, and deblockingartifacts are most observable in these regions.Secondly, a adaptive bilateral in-loop deblocking filter which combines adaptivein-loop filter (ALF) and adaptive bilateral filter (ABLF) is proposed. First, forbilateral filter, a adaptive bilateral filter is designed by selecting adaptiveparameters has optimized and the parameters based on the image characteristics. Second, combined adaptive bilateral filter with adaptive in-loop deblocking filter,ABLF+ALF instead. The ABLF is particularly suited to remove the blocking artifactswhereas the ALF is more targeted to reduce the SSD between the original andreconstructed frames. Therefore, it is appropriate to combine these two filters byselecting, for each imge block in the reconstructed frame, the type of filter binging thebest RD performance. Combining the two filters, balancing the two features canachieve better deblocking results. This method is characterized by,(i)The ABLF+ALFprovides the highest bitrat reductions and PSNR improvements when compared toboth NO-ALF and ALF methods.(ii)The typical video scene sequences, introducinghigher coding distortions easily, the higher bitrat reductions with the batter qualities.(iii)The sequences which contain several textured detail and, therefore, are moreexposed to blocking artifacts, get desired results.Two methods have their advantages, because of basing on the HM-9.0software,methods have high versatility and scalability. Through typical video sequences ofsubjective and objective tests, the results show that the two methods from differentangles can improve the coding efficiency, get a good improvement of the video, andachieve good qualities for deblocking.