User-assisted sparse stereo-video segmentation

摘要: Motion-based video segmentation has been studied for many years and remains challenging. Ill-posed problems must be solved when seeking a fully automated solution, so it is increasingly popular to maintain users in the processing loop by letting them set parameters or draw mattes guide process. When multiple-view videos, however, amount of user interaction should not proportional number views. In this paper we present novel sparse algorithm two-view stereoscopic videos that maintains temporal coherence view consistency throughout. We track feature points on both views with generic tracker analyse pairwise affinity temporally overlapping disjoint tracks, whereas existing similar techniques only exploit information available tracks overlap. The use stereo-disparity also allows our technique process jointly views, exhibiting good output. To make up lack high level understanding inherent techniques, allow refine output split-and-merge approach as obtain desired view-consistent over frames few clicks. several real examples illustrate versatility technique.