Detection of slow moving video objects using compound Markov random Field model

Often, moving object detection in a video sequence has been achieved a variant of temporal segmentation methods. For slow moving video objects, a temporal segmentation method fails to detect the objects. In this paper, we propose a Markov random Field (MRF) model based scheme to detect slow movements in a video sequence. The proposed scheme is a combination of a proposed spatio-temporal segmentation scheme and temporal segmentation method. A compound MRF model is used in spatio-temporal framework. In this framework, the a priori distribution is MRF and this takes care of spatial distribution of current frame, temporal frames and the Change Detection Masks (CDM) of the temporal frames. The spatio-temporal segmentation problem is formulated as a pixel labeling problem in Maximum a posteriori (MAP) framework. The MAP estimates are obtained using a hybrid algorithm. These estimated labels are used to obtain the Video Object Plane (VOP) and hence the detection of objects. The results are compared with joint segmentation scheme (JSEG). Results presented demonstrate that the proposed scheme with CDM model could detect slow moving video objects.