Applying Unsupervised Fuzzy C-Prototypes Clustering in Motion-Based Segmentation

Recent technology in digital video processing has moved to “content-based” storage and retrieval. There is a strong demand to segment multiple moving objects in a video sequence so that a higher level task can be performed on the individual object. In segmentation, robust clustering is a necessary step which plays an important role in identifying regions. In this paper, we present a scheme for extracting moving objects. First, the dense optical flow fields are calculated to extract motion vectors. Surface fitting is performed over the parametric motion model. Then, an unsupervised robust fuzzy C-Prototypes clustering technique is applied to motion-based segmentation in the parameter space. Finally, the individual moving object and background can be represented in layers. Experimental results showing the significance of the proposed method are provided.