Entropy classification and discrete-cosine-transform-based image indexing system

With the rapid growth of multimedia technology, more and more multimedia content is disseminated in networks or stored in databases. Image data is one of the multimedia types to be seen or accessed by users on the Internet or from databases. Searching the related images by querying image content is helpful for the management and usage of an image database. Therefore, research on image indexing techniques is an important topic. We propose an efficient content-based image retrieval (CBIR) system. Basically, the proposed method is a discrete cosine transform (DCT)-coefficient-based technique that extracts content features using some DCT coefficients. In addition, our method also uses entropy to classify images in the database so that it can reduce the search space to decrease the processing time. The proposed system has the property of robustness to rotation, translation, cropping, noise corruption, etc. The indexing time is only about 4 to 10% compared to most recently published results. According to our experiment results, the system is highly efficient in terms of robustness, precision, and a processing speed. (C) 2006 SPIE and IS&T.