基于选择特征宏块的快速视频稳像  

Fast video stabilization based on selecting featured macroblocks

作　　者：李沙[1] 谢晓明[1] 

机构地区：[1]北京化工大学信息技术与科学学院,北京100029

出　　处：《电子设计工程》2015年第24期183-185,共3页Electronic Design Engineering

摘　　要：针对抖动较小、前景较少的视频序列,提出了一种特征宏块的选择的快速视频稳像算法。所提方法中,根据视频帧宏块的内部相邻小宏块间的误差绝对值之和将宏块定义为有效宏块或无效宏块,选出有效宏块并进行全局运动估计,从而获得准确的全局运动矢量。通过实验验证,相较于未对宏块分类,该方法不仅可以有效地节省搜索时间,还能提高视频稳像的质量和精确性。According to video sequences with small jitters and less foreground, a novel algorithm is proposed for fast video stabilization using selection of featured macroblocks .In this proposed method , a macroblock is labeled as a valid one or a invalid one according to the sum of absolute difference between adjacent small macroblocks in some macroblock of a video frame sequence. After that, global motion estimation is performed by picking out valid macroblocks,thus, accurate global motion vectors are obtained .The experimental results show that video stabilization using the proposed method can not only reduce search time effectively , but also improve the quality and accuracy of video stabilization compared to unclassified macroblocks.

关 键 词：视频稳像 全局运动估计 有效宏块 全局运动矢量 

分 类 号：TN911.73[电子电信—通信与信息系统]