Affiliation possibility map filtering for image segmentation improvement

Image segmentation is an important task in computer vision. The task of image segmentation is to portion image into segments, thus provide more meaningful information of the image contents. Many methods have been developed for numerous application. The common problems of most of the segmentation techniques are scattered segmentation lines, too much details, small or thin segments, and noisy segmentation. In this paper, we propose a novel method for image segmentation improvement. We present state-of-the-art methods, our proposed method and experiments, showing performance over segmentations from classical image segmentation methods and also over segmentations from background subtraction methods.