Fast and Efficient Indexing and Similarity Searching in 2D/3D Image Databases

D-index is among the most efficient similarity search indexes, its performance dramatically depends on the choice of the ρ parameter. In This paper we propose a new criteria and a technique that ensures a good choice of D-index ρ parameter for reducing the D-index searching cost with an acceptable construction cost. We present our results upon two 2D/3D image databases; the Amsterdam Library of Images ALOI-1000 consists of 72000 color images of views and COIL-100 database consists of 7200 color images of views. We have described each view by combination of three well established descriptors from the MPEG-7 standard: CSD, SCD and EHD, and we have used the recommended distance measures by MPEG-7 to compare views. The results obtained prove the search efficiency of the proposed method against the original D-index and the sequential method