despite the great progress on interactive image segmentation, image co-segmentation, 2d and3dsegmentation, there is still no workable solution to the problem: given a set of calibrated or un-calibrated multi-view im...
详细信息
despite the great progress on interactive image segmentation, image co-segmentation, 2d and3dsegmentation, there is still no workable solution to the problem: given a set of calibrated or un-calibrated multi-view images (say, more than 40 images), by interactively cutting 3 similar to 4 images, can the foreground object of the rest images be quickly cutout automatically and accurately? In this paper, we propose a non-trivial engineering solution to this problem. Our basic idea is to integrate 3dsegmentation with 2dsegmentation so as to combine their advantages. Our proposed system iteratively performs 2d and3dsegmentation, where the 3dsegmentation results are used to initialize 2dsegmentation and ensure the silhouette consistency among different views and the 2dsegmentation results are used to provide more accurate cues for the 3dsegmentation. The experimental results show that the proposed system is able to generate highly accurate segmentation results, even for some challenging real-world multi-view image sequences, with a small amount of user input. (c) 2013 Elsevier Inc. All rights reserved.
暂无评论