基于运动矢量的运动物体提取方法

为提高视频运动物体提取的准确性，提出了一种新的基于运动矢量信息的视频运动物体提取方法。将与运动矢量信息相关的分割因子用于运动物体提取的判据，然后通过对每帧图像构建分割因子矩阵，对每一帧分割因子矩阵进行二维熵门限判决，实现对视频序列中的运动物体的提取。研究结果表明该算法对于全局运动比较稳定且运动物体的速度较慢的序列具有较好的提取效果，提取成功率达90%以上。

Abstract:

 In order to improve the accuracy of segmentation of motion object against the background，this paper proposed an algorithm of motion object segmentation based on motion vector information of macro block. The algorithm used motion vector information to generate segmentation factor，and constructed matrix of segmentation factor to each frame. Finally the 2D entropy threshold algorithm was chosen to judge the segmentation threshold in each frame，and completed the segmentation process of motion object in video sequences. The proposed algorithm has better extraction performance for the sequence which has stable global motion and slower speed of movement. The experimental results show that the successful rate of the motion object extraction is more than 90%.