基于图像块差分和Mean Shift算法的运动目标检测  

Motion detection based on image block difference and mean shift algorithm

作　　者：刘玉兰[1] 彭思龙[1] 

机构地区：[1]中国科学院自动化研究所国家专用集成电路设计工程技术研究中心,北京100080

出　　处：《计算机应用》2008年第8期2017-2020,共4页journal of Computer Applications

基　　金：国家"十一五"科技支撑计划资助项目(2006BAK07B04);中科院自动化所青年科技创新基金资助项目(08J1061GZ1)

摘　　要：运动目标检测是计算机视觉中的一个重要研究内容,现有算法中的一个重要问题是噪声对分割结果的影响。提出了一种时空域信息相结合的运动目标检测算法:首先利用图像块的重心位置在时间域上差分结果初始化目标轮廓,图像块差分的方法可以消除噪声的影响及减少目标内部的空洞;然后采用Mean Shift算法对初始轮廓进行迭代,使其逐步贴近真实的目标边缘。实验表明该算法能快速准确地分割出序列图像中的运动目标。Motion detection is an important research area in computer vision. A conventional problem in existing algorithms is the influence of noises on the segmental results. A new motion detection algorithm based on joint information in spatio-temporal domain was presented. First, result by temporal difference of image block~ barycenter was used to initialize an object contour. The image block difference algorithm can eliminate the influence of noise and reduce black holes in objects. Then, a mean shift algorithm was used to iterate the initial contour to make it close to the object step by step. Experiments show that this algorithm can detect moving objects in image sequence quickly and accurately.

关 键 词：视频监控 运动分割 图像块差分 偏移的均值向量 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]