论文部分内容阅读
为了降低高效视频编码(HEVC,high efficiency video coding)的帧间预测复杂度,提出了一种基于运动特征的HEVC快速帧间预测新方法。首先利用视频相邻帧的时域相关性,通过计算每个待编码单元(CU)及其子块的帧差离散度(FDD)确定该CU的最佳编码深度d;再依据该深度下CU的区域运动特征(RMFd)将待编码CU划分为3类运动区域,进而确定该CU的候选帧间预测模式,减少不必要的帧间预测模式遍历过程。试验结果表明,本算法可以在保证编码性能损失不大的前提下显著提高编码效率;与标准算法相比,在低延时和随机访问两种编码结构下,同等客观质量下码率(BDBR)分别增加0.89%和0.83%,同时节省了51.6%和48.5%的编码时间。
In order to reduce the inter prediction complexity of HEVC (high efficiency video coding), a new HEVC fast inter prediction method based on motion characteristics is proposed. Firstly, using the temporal correlation of adjacent frames in a video, the optimal coding depth d of the CU is determined by calculating the frame-by-frame disparity (FDD) of each unit to be coded (CU) and its sub-blocks. Based on the CU (RMFd) divides the to-be-coded CU into three types of motion regions to determine the candidate inter prediction mode of the CU and reduce unnecessary inter-prediction mode traversal process. The experimental results show that the proposed algorithm can significantly improve the coding efficiency without any loss of coding performance. Compared with the standard algorithm, under the two coding schemes of low latency and random access, the same objective quality coding rate (BDBR) Increased by 0.89% and 0.83% respectively, while saving 51.6% and 48.5% coding time.