User-assisted segmentation algorithm using B-spline curves
Daehee Kim, Yo‐Sung Ho · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2000
Most automatic segmentation techniques have difficulties in extracting individual video objects in a single frame. They are somewhat premature to obtain desirable segmentation results from various kinds of image sequences. However, if the user can provide semantic information of video objects for the first frame in a user-assisted manner, improved segmentation results can be obtained in the following picture frames. The user-assisted or semi-automatic approach for video segmentation is more practical in generating VOPs of moving objects. In this paper, we propose a new semi-automatic video segmentation scheme using an active contour algorithm. In most active contour algorithms, samples of the contour convey insufficient information about the curve shape between samples due to finite difference approximation. In addition, since most active contour algorithms have been developed for images of homogeneous simple background, they are not applicable to moving objects in the complex background. Therefore, this paper addresses the B-spline representation of curves and the morphological feature extraction for video object segmentation.