Research On Fast Mode Decision Algorithm In HEVC

In the past few decades, a series of video coding standards are developed, such asH.264/AVC and MPEG-2. These video coding standards are widely used in a varietyof domains. In recent years, videos with high definition and ultra-high definition arebecoming more and more popular, which means the amount of video data to be dealedwith are now becoming more huge. Traditional video coding standards are notperforming very well when dealing with those videos. So, VCEG(Video CodingExperts Group) and MPEG(Moving Picture Experts Group) established JCT-VC(JointCollaborative Team on Video Coding) to develop the next generation video codingstandard—HEVC(High Efficiency Video Coding).At present, although HEVC has not been finished, the whole coding structure hasalmost been established, and the compression performance of HEVC is nearly twiceas H.264/AVC standard. HEVC adopts tree coding structure, using flexible block sizeand up to35kinds of intra prediction modes. Those made the intra and interprediction of HEVC more accurate, but also increased the computational complexity.How to reduce the computational complexity of HEVC is becoming a popularissue now. This paper analyzed mode decision process in HEVC, and proposed fastintra and inter mode decision algorithms to reduce the encoding complexity of HEVCwhile maintained the coding efficiency. The major works of this paper are as follows:1) To reduce the computational complexity of intra prediction, a fast intraprediction mode selection algorithm is proposed. First, the spatial correlationsbetween Current CU (Coding Unit) and neighboring CUs are used to predict CurrentCU’s best encoding type. Then, In order to solve the problem that the search processof the best intra prediction mode is quite complicated, the best prediction direction inthe upper depth CU is used to rule out the most impossible candidate predictiondirections in current depth CU. Only prediction directions around the best predictiondirection of upper depth CU needed to be calculated. The simulation result shows the proposed fast intra prediction algorithm can save about28.89%encoding time anddecrease of PSNR (Peak Signal to Noise Ratio) is0.07dB.2) The inter prediction procedure of HEVC is also optimized in this paper. Twofast inter prediction algorithms are proposed to reduce the inter prediction complexityof HEVC.(1) The first algorithm using ASSD (Average Sum of Square Difference)between current CU and Co-Located CU to decide whether current CU is in stationaryarea or motion area. If current CU is in stationary area, no further split is performed, ifcurrent CU is in motion area, then all the prediction modes in current are skiped. Thesimulation result indicates that this method saves the encoding time by50.88%.(2) The latter algorithm uses the correlation between current CU and itsspatial-temporal neighboring CUs to skip the calculation of certain depth CU andprediction modes. The simulation result indicates that this method saves the encodingtime by51.91%with drop of PSNR by0.04dB.