Research On Fast Inter Coding Algorithm For HEVC

With the rapid development of the application of High Definition(HD) video, there is a huge challenge to its transmission and storage. In the new environment, the Joint Collaborative Team on Video Coding(JCT-VC) developes a new video coding standard namely High Efficiency Video Coding(HEVC/H.265) to deal with the challenge. To achieve the high efficiency of video coding, HEVC addes lots of new advanced video coding technologies such as flexible block partition, asymmetric motion partition, multi prediction block mode on the basis of the original video coding standard(AVC/H.264). Compared to H.264, HEVC can save about 50% bit rate while maintaining the same subjective video quality and the complexity of the encoding computation for HEVC is also greatly increased at same time. Because of the low computing capacity of wireless mobile devices, they cannot support the use of all the coding technologies, which is bound to greatly reduce the coding efficiency, and thus cannot highlight the advantages of coding efficiency of HEVC. The inter prediction is one of the most important part of prediction coding, which is the most complex part in video coding. Therefore, the fast coding algorithm of HEVC is one of the most important issue in the field of video coding.This paper gives an introduction to the development and the whole framework of video coding standard in brief at first. Secondly, it summaries the research on fast coding for HEVC at home and abroad systematically. Then, it gives a short introduction on the technical framework and key technologies of HEVC. Finally, CU(Coding Unit) block partion and PU(Prediction Unit) mode decision for HEVC are introduced in detail and deeply studied.In the process of HEVC CU block partion, the current LCU(Largest Coding Unit) draws a depth range by using the information of adjacent LCU’s maximum depth. By this way, the process of PU mode decision under a few depths can be skipped and the coding complexity is significantly reduced. In order to further reduce the computational complexity of CU partion, an early CU termination algorithm is put forward after PU mode decision at each depth according to the residual information and transform quantitative coefficient matrix.In HEVC inter PU mode decision process, fast skip algorithm and fast 2Nx2 N algorithm are proposed in this paper. Both skip mode and 2Nx2 N mode is accounted for largest in PU mode decision. In order to predict the Skip mode early, Skip mode decision is put at the beginning of the process of PU mode decision. What’ more, the fast skip algorithm is put forward by mining the relationship between the residual error information, transform coefficient, quantization parameter and skip mode. The fast 2Nx2 N algorithm is proposed by analyzing the relationship between motion vector, number of nonzero transform-quantization coefficient, TU(Tranform Unit) depth and 2Nx2 N mode. The experimental results show that the two fast mode decision algorithms have obtained very ideal coding results. As the two algorithms aimes at two different PU mode, they can be consolidated into use. Under the condition of combination of two algorithms, the coding time can be reduced about 44% under the setting of Low Delay_P and 51% under the setting of RandomAccess on the basis of video quality changes little.By making good use of the coding information and residual information in video image, the proposed algorithms can make accurate judgment for CU block partition and PU mode decision early, and thus it can speed up the HEVC encoder considerably. As a result, the algorithm has a certain reference value for application of HEVC real-time system in the future.