A method of the moving object region clustering in video sequence based on the information bottleneck theory is realized in this paper.The clustering principles and characteristics of the information bottleneck method are presented.All the images in video sequences are changed with the method of Gaussian mixture modeling to form sub-regions,and then the various sub-region clustering is carried out according to information bottleneck method to make the mutual information loss least.The experimental results show that the clustering method can achieve the moving target region clustering effectively.