摘要
为了减少运动估计的计算复杂度并提高其搜索性能,提出了一种基于低频子图的运动估计算法。该算法首先将当前帧和参考帧通过低通滤波器得到数据量减为1/4的低频子图,然后用全搜索得到最佳匹配的低频子块,最后对低频子块所覆盖的原参考帧区域进行精细搜索得到最优点。实验结果表明,与传统运动估计算法相比,该算法提高了运动估计的精确度和降低了计算复杂度,并且对各类视频都有很高的鲁棒性。
In order to reduce the computational complexity of motion estimation and improve the accuracy, a fast motion estimation algorithm based on low frequency sub-image was proposed. The algorithm firstly puts Current Frame and Reference Frame through a low-pass filter to get low frequency sub-images with quarter pixels, then attains the best matched sub-block by Full Search, at last the area of Reference Frame covered by the sub-block are searched to get the best matched point. Compared with traditional algorithm, experimental results show that this algorithm gives a significant improvement in accuracy for motion estimation and reduces the computational complexity, and nossesses strong robustness in different kinds of video sequences.
出处
《计算机应用》
CSCD
北大核心
2006年第5期1024-1026,共3页
journal of Computer Applications
基金
国家自然科学基金资助项目(60372057)
关键词
低频子图
运动估计
HAAR小波
块匹配算法
视频编码
low frequency sub-image
motion estimation
Haar wavelet
block matching algorithm
video coding
