Abstract: Based on DSP characteristics and motion estimation’s accuracy and efficiency requirements, a new adaptive search-window fast motion estimation search algorithm is proposed. This algorithm uses the spatial and temporal correlations among video images and the statistical characteristics of motion vectors, adaptively adjusts the search-window on frame layer and block layer, and searches the window by using diamond search (DS) method. Experimental results show that, on average, it saves 50% data transmission compared with fixed-size search-window algorithm and 3 points compared with DS, maintaining high accuracy of estimation. And it suits for different types of video sequences adaptively.

Key words: motion estimation, adaptive search-window, PSNR, search points, block match

摘要： 针对DSP芯片特点及运动估计精度和效率的要求，提出一种新的自适应搜索窗口快速运动估计搜索算法。该算法充分利用视频图像的时间、空间上的相关性以及运动向量的统计特性，在帧层和块层自适应调整搜索窗口，利用钻石搜索法进行搜索。经实验验证，该算法在保证运动估计精度的情况下，数据传输量比固定搜索窗口算法平均少50%，搜索点数比钻石法平均少3个，对不同类型视频序列都有较好的自适应能力。

关键词: 运动估计, 自适应搜索窗口, 峰值信噪比, 搜索点数, 块匹配

CLC Number: 

• TP391