According to the author, this book is meant for researchers and practitioners concerned with digital libraries and image database systems, and can also be used as a textbook or reference. It provides some brief notes on basics, emphasizing the importance of the choice of image features as well as the difficulty of bridging the gap between features that can be captured and content-based image retrieval. Chapter 2 is a survey of the state of the art, focusing on example systems, such as QBIC, which are described and compared in some detail. Chapter 3 addresses data management, illustrating index structures and their use in search and image comparison.
Chapters 4 through 8 are devoted to a detailed account of the ARBIRS system. ARBIRS is designed to retrieve images with specific shapes, as well as those related to a given image. It is therefore based on identifying regions. Images are characterized by object regions defined by segmentation, and texture regions represented by histograms. Images are retrieved by simple or compound object region and texture region queries. All this is described in detail, especially the crucial segmentation subsystem. Finally, chapter 9 discusses implementation and evaluation with a sample image database containing 50 image classes, measuring success in retrieving images from the right class, and only those, for different query types.
The book serves its purpose as a useful snapshot of a current approach and system. It is clearly written, with detail suited to classroom work; some mathematical background is needed. It seems to be too early to judge the potential of ARBIRS’s novel region-based approach.