文章基本信息

标题：Fast Motion Estimation for Quad-Tree Based Video Coder Using Normalized Cross-Correlation Measure
本地全文：下载
作者：Dr. Eskinder Anteneh Ayele ; Mr. Ravindra Eknath Chaudhari ; Dr. S. B. Dhok 等
期刊名称：International Journal of Image Processing (IJIP)
电子版ISSN：1985-2304
出版年度：2013
卷号：7
期号：4
页码：330-338
出版社：Computer Science Journals
摘要：Motion estimation is the most challenging and time consuming stage in block based video codec. To reduce the computation time, many fast motion estimation algorithms were proposed and implemented. This paper proposes a quad-tree based Normalized Cross Correlation (NCC) measure for obtaining estimates of inter-frame motion. The measure operates in frequency domain using FFT algorithm as the similarity measure with an exhaustive full search in region of interest. NCC is a more suitable similarity measure than Sum of Absolute Difference (SAD) for reducing the temporal redundancy in video compression since we can attain flatter residual after motion compensation. The degrees of homogeneous and stationery regions are determined by selecting suitable initial fixed threshold for block partitioning. An experimental result of the proposed method shows that actual numbers of motion vectors are significantly less compared to existing methods with marginal effect on the quality of reconstructed frame. It also gives higher speed up ratio for both fixed block and quad-tree based motion estimation methods.
关键词：FFT; Motion Estimation; Normalized Cross Correlation; Quad-tree; Video Compression.