Efficient block-based matching for content-based retrieval of CT head images

An efficient image retrieval system for use in medical applications is proposed. The system addresses both global and local features and is able to perform similarity retrieval on the whole image as well as on local regions within the image. The retrieval algorithm is based on a simple grey level histogram (GLH) algorithm to capture the intensity distribution, and a novel block-based retrieval technique to localize the image. The algorithm is tested on a database of computed tomography (CT) images of head provided by a local hospital. Promising results are reported which pave the way for the integration of the block-based method with a more advanced feature extraction method.