Semiautomatic segmentation using an edge map

Abstract

A semiautomatic segmentation method has been developed using an edge map. Although the ideal segmentation method is a fully automatic segmentation without user interation, current technologies for automatic segmentation is somehow premature to obtain desirable segmentation results about all kinds of image sequences because the mathematical model or the similarity measure for the extraction of video object can not be objectively defined. To solve this problem, the semiautomatic segmentation, i.e., user-assisted segmentation, has been used currently. A user can initially draw the object boundaries. Then, the segmentation algorithm is applied to the image using the initial contour drawn by user. This thesis proposes a segmentation method that finds the exact boundaries by using edge map. The edge map is composed of the weighted summation of a contour drawn by user, gradient image, and Canny``s edge image. We can extract the user- defined object by using the edge map iteratively. The iteration makes the initial contour drawn by user approach the extract object. The proposed method can be used not only for intra-frame segmentation but also inter-frame segmentation.