Skip to Main Content
We present a new matching method called tuple matching (TM), which is an algorithm for matching image signatures. Since signatures can contain arbitrary features like color, shape, and texture, we focus on signatures that are generated from color histograms by using graph theoretical clustering (GT-clustering). In contrast to histogram intersection (HI) (Swain, M.J. and Ballard, D.H, 1991) or similar approaches, TM defines a similarity measurement with a many to many mapping between tuples in an arbitrary neighborhood in spite of using a one to one mapping between bins as defined by HI. As a result, TM is more robust than HI when the illumination is changing. In contrast to earth mover's distance (EMD) (Rubner, L.J.G.Y. and Tomasi, C., 1998), similarity between signatures is not calculated by using a solution of the transportation problem. Thus the performance of TM is better than EMD.