图像序列中运动人体的分割方法的研究

Study on segmentation of motion body in image sequence

在对现有的运动分割算法的研究基础上结合运动分析方法,对基于图像帧差法构造背景图像进行了深入的分析和研究,提出了一种利用阈值来判定人体信息和背景信息以达到背景减除的新算法来实时地提取出运动的人体,在利用帧差法提取出背景图像后,通过当前帧和背景图像的像素值的对比减除可快速精确地完成对运动人体的分割。在NLPR数据库中进行了实验,结果表明本算法能快速有效地一次提取出多个人体目标,并且失真度比较小。

Together with the motion analysis methods and on the basis of gaining the background image by using the change detection in image sequence method, a new algorithm, which could get the motion bodies in real time by distinguishing the background information and the body information through a threshold, is proposed after full analysis. After getting the background frame by the CDM, the bodies could be separated by subtracting the background information from the current frame＇s and it could achieve the segmentation fleetly and accurately. Extensive experimental results based on the NLPR Database demonstrated that the proposed algorithm could get several motion bodies one time with relatively lower computational cost and little distortion.