In this paper, region based and location based retrieval systems have been implemented for retrieval of MR-T2 axial 2-D brain images. This is done by extracting and characterizing the tumor portion of 2-D brain slices by use of a suitable threshold computed over the entire image. Indexing and retrieval is then performed by computing texture features based on gray-tone spatial-dependence matrix of segmented regions. A Hash structure is used to index all images. A combined index is adopted to point to all similar images in terms of the texture features. At query time, only those images that are in the same hash bucket as those of the queried image are compared for similarity, thus reducing the search space and time.