Similarity Matching Of Arbitrarily Shaped Video By

Abstract

The increasing availability of object-based video content requires new technologies for automatically extracting and matching of the low level features of arbitrarily shaped video. In this paper, we propose a shape retrieval method for video objects. Our method takes into account not only the still shape features but also the shape deformations that may occur in the lifespan of video objects. We define a new shape similarity measure that is based on the shape similarity of the representative temporal instances of video objects. We also propose shape deformation features that are based on the variances of the still shape features. The proposed visual features can be derived directly from the MPEG-4 compressed domain or computed from the shape masks of the video objects in the spatial domain. Our experiments show that the proposed method offers good retrieval performance results.