Description

Despite the fact that there have been a lot of past work on video segmentation technique, it is still a difficult task to extricate the video objects precisely without interactions, particularly for those videos which contain irrelevant frames (frames containing no normal targets). In this paper, a novel multi-video object co-segmentation technique is raised to co-segment normal or comparable objects of pertinent casings in various videos, which incorporates three stages: 1) object proposal generation and clustering within each video; 2) weighted graph construction and common objects selection; and 3) irrelevant frames detection and pixel-level segmentation refinement.