An image database system with content capturing and fast image indexing abilities

Abstract

While general object recognition is difficult, it is relatively easy to capture some primitive image properties such as color distribution, prominent regions and their geometrical properties from an image and use these features to narrow down the search space when attempts to retrieving images by contents from an image database are made. The authors are building an image database in which images are indexed by both the numerical index keys generated automatically from the captured primitive image features using a set of rules, and traditional descriptive keywords entered by users when images are loaded. Users can either use the descriptive keywords or provide information regarding these image properties to query and retrieve images from the database. With this approach, the authors turn the difficult problem of image matching into image retrieval by index keys, which can be performed easily and rapidly using current database techniques. Initial experiments on the image database system show promising performance.