Using Domain Knowledge in Querying Image Databases.
Tat‐Seng Chua, Kah-Choo Teo, Beng Chin Ooi, Kian‐Lee Tan · 1996
Introduction The advances in computer technology enable us to store and manipulate a massive amount of color images on the computer. Databases containing large on-line image collections are becoming common. Current and potential applications for imaging and other media include: medicine; photo-journalism; marketing and promotion; education and training; and entertainment. The emergence of large and complex image databases has led to the demand for tools that can manage, organize and retrieve images effectively. Most early image retrieval systems use keywords or free-text descriptions supplied by the authors as the basis for retrieval [1, 14]. The problem with text-based retrieval systems is that the text descriptions must be entered manually by the authors, and hence the descriptions tend to be incomplete and inconsistent. Because of this problem, recent systems [5, 6, 8] use the visual attributes of images, such as colors, textures, and shapes, as indices for retrie