A MPEG-7-aided segmentation tool for content-based video coding

A segmentation algorithm for content-based video coding is presented. The fundamental assumption of our approach is that many video scenes´ content can be classified into subjectively relevant and subjectively irrelevant textures. The texture categorization is thereby done by a texture analyzer (encoder side), while the corresponding texture synthesizer assumes the replacement of the subjectively irrelevant textures (decoder side), given the side information generated by the texture analyzer. The focus of this paper is the extensive evaluation of the texture analyzer, which uses MPEG-7 descriptors for similarity estimation. The texture analyzer is based on a split and merges segmentation approach and also provides solutions concerning temporal mapping of the identified subjectively irrelevant textures. The current implementation of the texture analyzer yields an identification rate of up to 93%.