Image Retrieval by Content Measure Metadata Coding

Many image collections contain few or no index terms. To search these collections, a battery of techniques known as Content Based Image Retrieval (CBIR) is used. With CBIR, the user need identify only an image exemplar. The exemplar is processed to extract content measures such as greyscale, lines, and edges. These measures are matched against a database of identical measures for a collection and the results returned to the user. The matching process between image search example and stored image content measures is complex and requires sophisticated data management support. In this paper, image content measures are encoded as metadata tags for the images, and included in the html image header. The support requirements are eliminated. Searching is simplified. Image content index measures are conveniently carried along with the image in an html wrapper. Collections demonstrating the technique are WWW available. Image processing and search software is available without copyright or fee from the author.