A universal scheme for content-based video representation and indexing

Abstract

This paper presents a universal scheme for automatic content-based video representation and indexing. Firstly, a video sequence is subdivided in time into shots each of which has similar content. These shots can be represented by key frames. Key frames provide a suitable abstraction and framework for video representation. Secondly, we could support an even wider range of video representation and indexing by segmenting and extracting smaller units (key object) within the key frame. In this paper, we also propose a novel approach of object segmentation.