Computer Engineering and Applications ›› 2007, Vol. 43 ›› Issue (14): 56-59.

• 学术探讨 • Previous Articles     Next Articles

Research on Motion Segmentation By Integrating Maximizer of the Posterior Marginals with MAP

Shitong Wang   

  1. (School of Information, Southern Yangtze University, Wuxi Jiangsu 214122,China)
  • Received:2006-06-12 Revised:1900-01-01 Online:2007-05-10 Published:2007-05-10


令狐永芳 郭小猛 王士同   

  1. 江南大学 信息工程学院 江南大学信息工程学院
  • 通讯作者: 令狐永芳

Abstract: A novel video motion object automatic segmentation algorithm based on Gaussian Markov random field is studied in this paper. In this algorithm, the probability density functions of the differente images are estimated as Gaussian mixture distributions. A fast estimation procedure for the posterior marginals is added to the MAP algorithm.Firstly,initial segmentation is applied to obtain number of the initial motions and the corresponding initial parameters of the motion model.Then the parameters are updated by using the given parameter estimation method. The experiment results show that the proposed algorithm here is effective.

Key words: Markov Random Field, Moving object segmentation, Maximizer of the Posterior Marginals, MAP algorithm

摘要: 本文探讨了一种基于高斯马尔可夫随机场(GMRF)模型的运动目标自动分割算法.该算法采用高斯混合分布描述视频序列的差分图像,对标准MAP算法进行了改进,使用快速方法计算后验边缘。首先对视频处理对象进行初始分割,获取初始运动数目以及相应的运动模型的初始参数,然后通过参数估计,不断更新模型参数,之后通过把每个运动区域和运动模型相关联,来同时估计多个运动区域,最终达到分割的目的.实验结果证明,本文所提的方法对运动目标分割具有较好的分割效果.

关键词: MRF 视频对象 最大后验边缘概率 MAP算法