Classifier Combination for Improved Motion Segmentation
Multiple classifiers have shown capability to improve performance in pattern recognition. This process can improve the overall accuracy of the system by using an optimal decision criteria. In this paper we propose an approach using a weighted benevolent fusion strategy to combine two state of the art pixel based motion classifiers. Tests on outdoor and indoor sequences confirm the efficacy of this approach. The new algorithm can successfully identify and remove shadows and highlights with improved moving-object segmentation. A process to optimise shadow removal is introduced to remove shadows and distinguish them from motion pixels. A particular advantage of our evaluation is that it is the first approach that compares foreground/background labelling with results obtained from ground truth labelling.