Research Of Fast Depth Decision Algorithm Of Inter Prediction For HEVC

With the development of multimedia technology and the expansion of network bandwidth, people on the video quality of the increasingly high demand, and the resolution of 2K 4K HD video appeared, it also brings a great challenge to the video transmission and storage. The joint group of experts (Joint Collaborative encoding Team on Video Coding, JCT-VC) released a new generation video coding standard HEVC (High Efficiency Video Coding, HEVC) compared to the previous generation standard H26 codec 4/AVC, the encoding unit from the original 8x8 extended to 64x64, at the same time in the process of encoding includes encoding tree (Coding Unit Tree, CTU unit), intra prediction mode selection, loop filter and other aspects of the greatly improved. Multi mode and multi depth traversal calculation in high prediction accuracy at the same time, but also greatly increase the computational complexity, it is difficult to apply to the actual.In view of several inter prediction statistical information similarity algorithm and the complexity of the parameters are difficult to obtain, the actual operation is difficult to achieve complex shortcomings, put forward simple and intuitive, can effectively block adjacent performance depth weighted similarity weighted prediction algorithm. The algorithm of depth by traversing the collection is currently CTU encoding the left and left on the forecast, and recently the reference frame near CTU 4x4 block maximum depth weighted evaluation, obtain the current predicted depth encoding CTU. By combining a variety of weighting coefficients of HM encoding comparison, and greater than the time weighted coefficient in space weighted coefficients are equal, and the coefficients of 1, plus The algorithm can achieve the best effect prediction right depth. Due to the non integer prediction depth, this paper can clearly show the encoding CTU neighboring blocks and its movement intensity was 0.5 based on the classification of depth category partition boundaries, get the current actual encoding depth encoding CTU, ahead of the end in the iteration or retention of 2Nx2N encoding results, skip unnecessary depth, in order to reduce the computational complexity.Experiments show that the weighted depth prediction algorithm is simple, effective and practical, and it can reduce the computational complexity of the video quality at the same time with the loss of 35%.