A content-based-image-retrieval approach for medical image repositories

Medical image repositories present a challenging environment for query and search algorithms, due to the highly specialised nature of the images. This paper describes an approach to assessing image similarity using simple statistical measures computed over blocks of adjacent pixels. The approach can be applied as an initial step in construction of a content-based-image-retrieval system for large collections of medical images, to rank candidate images according to their similarity to a given query image. Experimental results demonstrating the effects of different choices of the block parameters are presented to indicate the robustness of the approach