Enhancing foreground segmentation by motion-based contour

Abstract

Foreground segmentation has been widely used in many computer vision applications, and background substraction(BS) is one of most widely used routines. In order to achieve satisfactory completeness and robustness of foreground, in this paper, we propose a method which uses motion-based contours to enhance the existing kinds of BS algorithms in the area which is low contrast. First we obtain key regions by motion-based contours and superpixels. Secondly, we use Otsu algorithm to determine the threshold in these areas. Besides that we take the advantage of feedback mechanism and the attributions of integral image to reduce the light spots in the foreground images. Experiments demonstrate that the proposed method can improve the completeness of foreground segmentation.