基于分类的运动矢量平滑和纠正方法  

A motion vector smoothing and correction method based on classification

作　　者：张为亚[1] 

机构地区：[1]上海交通大学上海数字媒体处理与传输重点实验室,上海200240

出　　处：《信息技术》2014年第9期93-97,101,共6页Information Technology

摘　　要：对于帧率上变换算法中运动矢量估计不准确的问题,文中提出了一种新的运动矢量处理算法,较好地纠正了错误的运动矢量。该算法根据前后向运动矢量的一致性程度,以及对应块的绝对差值和,对运动矢量进行了可靠性分类。针对分布在纹理相似区的错误运动矢量,利用时间和空间相关性联合纠正;对于物体边缘区域则分解为小块重搜索运动矢量,以避免边缘模糊现象。实验结果显示,算法可以较好地纠正运动矢量。In this paper, a novel motion vector processing method is proposed to detect, classify and correct those unreliable motion vectors in frame rate up-conversion. Firstly, it obtains motion vectors reliability information through the consistency of the forward and backward motion vectors and the sum of absolute difference of the corresponding block. For those motion vectors distributed in the texture similar area, it corrects them by the time and spatial correlation. For the object edge region, it researches motion vectors by decomposed to small blocks. The experimental results show that the proposed method can correct motion vectors very well.

关 键 词：帧率上变换 运动估计 运动矢量分类 运动矢量平滑 

分 类 号：TN911.73[电子电信—通信与信息系统]