Multiple-Feature Spatiotemporal Segmentation of Moving Sequences using a Rule-based Approach

In this paper a novel two-stage architecture for object-based segmentation of moving sequences is proposed using multiple features such as motion, intensity and texture. The first stage locates perceptually meaningful objects using a hierarchy of single-feature segmentation processes. The second stage refines the boundaries of located objects using a combination of features according to a set of appropriate rules. Experimental results show that the proposed approach yields intuitively correct as well as accurate segmentations of moving sequences, which compare favourably with established state-of-the art techniques in the literature.