基于视频压缩的多分辨率预测菱形搜索算法

Multi-resolution predictive diamond search algorithm on video compression

为了减小视频压缩编码标准中运动估计算法的计算复杂度和提高运动补偿的准确性,考虑到视频图像质量和算法运行时间两者之间的关系,提出一种多分辨率预测菱形运动估计搜索算法,利用同一视频图像中的相邻宏块运动矢量相似的特征,预测当前块的搜索起点,采用大小不同的搜索匹配宏块,减少搜索范围和搜索点数目。实验结果表明,该算法与经典菱形搜索算法相比,搜索时间平均减少了0.5ms,信噪比平均提高了0.5dB。

In order to reduce the computational complexity of the motion estimation algorithm and improve the accuracy of motion compensation,a multi-resolution predictive diamond search algorithm is proposed for motion estimation algorithm on video compression standards,owing to the relationship between the quality of video and the running time of algorithm,using the comparability of motion vector in the same image and forecasting the preliminary point,reducing the searching ranges and numbers. Compared with diamond search algorithm,experimental results show that the time of motion estimation is reduced to 0.5 ms and PSNR is increased about 0.5 dB by the multi-resolution predictive diamond search algorithm.