Description:
Visual Information Retrieval, written by the head of the Systems and Information Department at the University of Florence in Italy, is a must-have for anyone developing systems for cataloging, searching, and retrieving visual-based data. A growing percentage of stored computer information is visual media, whether it's still images or video clips. This kind of data has a very different format and structure than text-based data, and cataloging, sorting, and searching non-text-based data presents a tremendous challenge to the contemporary database programmer. This is not a book for the casual programmer. It offers high-level suggestions on how to build the architecture for such systems, key theories on how to represent this kind of visual content, how to build similarity models, and indexing methods. It continues with examples of how to index and catalog still images based on color, texture, shape, and spatial relationship similarity. Chapter 6 details the problems and current solutions for content-based video retrieval. Detecting sharp transitions, how to analyze compressed and uncompressed streams, and gradual transition detection are just a few of the problems presented. The solutions presented are practical and fascinating. Visual Information Retrieval is a clearly written, although sometimes dense, handbook. The author uses a generous amount of examples of mathematical formulas, illustrations, and color plates. Clearly not a book for every database programmer--but a mandatory reference book for anyone building visual retrieval systems. --Mike Caputo
| 
