Techniques and index structures for efficient image retrieval on the web
Abstract
The rapid growth of images and photographs available on the Web today has created an emerging need for efficient retrieval techniques for visual information. Image searching on the Web today is carried out using text-based search methods. Searching is done by entering keywords describing the images to be retrieved. Not all aspects of an image are easy to describe by text. For instance, is difficult to describe the shape of a car by words only. Content-based image retrieval is another approach to image searching where the retrieval is based on the visual features found in the images and not on any external description. Content-based retrieval has been a active field of research for several years, but has not been used in any commercial Web search engines so far. This thesis presents techniques and index structures for building an efficient, content-based image search engine for the Web. The thesis presents requirements and issues which must be met, and provides a system design designated from the techniques presented. A prototype has been implemented for a demonstration of the concept.