Direct DCT indexing using genetic algorithm concepts

It is highly desirable in terms of speed and computational costs to perform image indexing and retrieval in the compressed domain. The exponential growth of digital media on both the WWW (Reddy and Fletcher, 1998) and home imaging equipment has prompted the development of faster, more accurate indexing algorithms. Successful techniques have the ability to summarise the features of an image into a relatively small key. We present a technique that uses the features of a genetic algorithm to quantize the content of an image directly in the DCT domain. The benefits of working in this domain, apart from the substantial savings of avoiding image reconstruction, is that the coefficients provide spatial information in a form well suited for image indexing.