Data and signal modeling for images and video sequences is experiencing important developments. Part of this evolution is due to the need to support a large number of new multimedia services. Traditionally, digital images were represented as rectangular arrays of pixels and digital video was seen as a continuous flow of digital images. New multimedia applications and services imply a representation that is closer to the real world or, at least, that takes into account part of the process that has created the digital information. Content-based compression and indexing are two typical examples of applications where new modeling strategies and processing tools are necessary: For content-based image or video compression, the representation based on an array of pixels is not appropriate if one wants to be able to act on objects, to encode differently the areas of interest, or to assign different behaviors to the entities represented in the image. In these applications, the notion of object is essential. As a consequence, the data modeling has to include, for example, regions of arbitrary shapes to represent objects. Content-based indexing applications are also facing the same kind of challenges. For instance, the video representation based on a flow of frames is inadequate for many video indexing applications. Among the large set of functionalities involved in a retrieval application, let us consider browsing. The browsing functionality should go far beyond the “fast forward” and “fast reverse” allowed by VCRs. One would like to have access to a table of contents of the video and to be able to jump from one item to another. This kind of functionality implies at least a structuring of the video in terms of individual shots and scenes. Of course, indexing and retrieval involve also a structuring of the data in terms of objects, regions, semantic notions, etc.