“Video segmentation with motion smoothness” by Wen, Wong, Chen and Sato
Conference:
Type(s):
Title:
- Video segmentation with motion smoothness
Presenter(s)/Author(s):
Abstract:
In this extended abstract, we propose a novel approach for video segmentation by utilizing motion information. Recently, graph-cutbased segmentation methods became popular in this domain but most of them dealt with color information only. Those methods possibly fail if there are regions similar in color between foreground and background. Unfortunately, it is usually hard to avoid, especially when objects are filmed under a natural environment. For instance, Figure 1(a) shows a result of graph cut with a small smoothness weighting, and hence some background regions are incorrectly labeled. On the contrary, if a larger smoothness weighting is used, some background regions near the foreground will be merged as shown in Figure 1(b). To improve those drawbacks, we propose a method based on both of color and motion information to conduct the segmentation. The method is useful because foreground and background usually have different motion patterns as shown in Figure 1(c).
References:
1. Brox, T., Bruhn, A., Papenberg, N., and Weickert, J. 2004. High accuracy optical flow estimation based on a theory for warping. In Proceedings of 2004 European Conference on Computer Vision, 25–36.
2. Li, Y., Sun, J., and Shum, H.-Y. 2005. Video object cut and paste. ACM Transactions on Graphics 24, 3, 595–600. (SIGGRAPH 2005 Conference Proceedings).
Additional Images:
