An image database system with fast image indexing capability based on color histograms

Abstract

While general object recognition is difficult, it is relatively easy to capture some inherent image properties, such as color distribution, to narrow down the search range when an attempt to retrieve images from an image database is made. We have built an image database in which images are indexed automatically using a histogram indexing method that we have proposed. With this method, a numerical index key is created for each image histogram using a set of mathematical formulas. By using this approach, we have transformed the difficult problem of image matching into one of image retrieval by index keys, which can be supported by existing database implementation techniques. We have also provided two query methods: image retrieval by user-provided sample images, and image retrieval by combinations of system-provided templates (such as sky, lawn, brick wall, and so on). Our initial experiments on the image database system have shown favorable results.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>