内容自适应的低复杂度多参考帧快速选择算法

针对移动视频业务的市场需求和移动终端设备的性能局限,提出了一种基于时空内容自适应的低复杂度多参考帧快速选择算法。在对视频序列编码模式采用率和参考帧时空相关性统计分析的基础上,根据时空相邻块最优参考帧的分布情况等信息,排除可能性较小的参考帧,缩小当前编码块的候选参考帧范围。并设定自适应的参考帧选择代价阈值,以判断是否需要扩大搜索范围。实验结果表明,该算法能大幅降低H.264的编码计算量,平均可降低49.12%的编码时间,而且基本保持编码率失真性能不损失。

Low-complexity Fast Multi-reference Frame Selection Algorithm based on content-adaptively

According to the market demand for mobile video services and the performance limited by the mobile terminal, a fast multi-reference frame selection algorithm which has low-complexity and content-adaptively is proposed. First of all, statistics the adoption rate of the best mode, and the spatial-temporal correlation of the best reference frame is analyzed. Then, the candidate reference frame is determined according to the best reference frame used in the temporal-spatial neighbors, exclude some less possible reference frame, narrowing the scope of the candidate reference frame. Furthermore, an adaptive threshold of reference frame selection based on the R-D cost is set, to determine whether need to expand the search scope. Experimental results demonstrate that the proposed algorithm can reduce the encoding time of 49.12% on average, while in keeping with the R-D performance almost unchanged.