Video segmentation based-on fuzzy clustering and multi-features

In this paper, we present a video segmentation algorithm for object-based coding based on fuzzy clustering using multi-features. Three features (color, motion and position) are chosen to consist of feature space, as inputs to the fuzzy clustering algorithm. A new block matching motion estimation algorithm is introduced and the value of DFD (displace frame difference) is used to measure the reliability of different features. At the same time, the distance metric and objective function of the fuzzy clustering algorithm are modified to improve the spatial cohesion of the segmentation results. Simulation results demonstrate the performance of the algorithm.