Collaborative Xmeans-EM clustering for automatic detection and segmentation of moving objects in video

Detecting and segmenting moving objects in video is a challenging and essential task in a number of applications. This paper presents a new algorithm of moving objects detection and segmentation. Firstly, we extract Selective Spatio-Temporal Interest Points (SSTIPs). The next step is to partition the SSTIPs into a set of moving clusters. To reduce the impact of the choice of a clustering method and its parameters on the quality of the result, we propose to integrate the concept of collaborative clustering of two clustering algorithms without requiring a user-defined number of clusters: Xmeans and Expectation-Maximization (EM) clustering. Finally, the segmentation of the objects associated to the given clusters is performed using an automatic maximal similarity based region merging (MSRM) method. Our algorithm is evaluated on several sequences and experimental results show a good performance for automatic detection and segmentation of moving objects.