Efficient similarity-based retrieval of temporal structures

Multimedia documents are composed of a multitude of objects of different media types. These different objects are related to one another by spatial and temporal specifications. In this paper we describe a framework for supporting efficient similarity based retrieval of temporal structures. We introduce the notion of similarity and the need for the same as well as discuss the properties of the similarity measure we propose. We discuss the development of similarity metrics to compare the temporal structures of multimedia documents. We also develop algorithms that efficiently index temporal structures based on these measures. More specifically, we propose a variant of the Multi Dimensional Scaling algorithm and show that the proposed variant is both efficient and provides high quality retrieval results.