Efficient similarity search using the Earth Mover's Distance for large multimedia databases

Abstract

Multimedia similarity search in large databases requires efficient query processing. The Earth Mover's Distance, introduced in computer vision, is successfully used as a similarity model in a number of small-scale applications. Its computational complexity hindered its adoption in large multimedia databases.We enable directly indexing the Earth Mover's Distance in structures such as the R-tree and the VA-file by providing the accurate 'MinDist' function to any bounding rectangle in the index. We exploit the computational structure of the new MinDist to derive a new lower bound for the EMD MinDist which is assembled from quantized partial solutions yielding very fast query processing times. We prove completeness of our approach in a multistep scheme. Extensive experiments on real world data demonstrate the high efficiency.

Parameterized earth mover's distance for efficient metric space indexing

“…The Earth Mover's Distance is defined between two feature signatures S q and S o as a minimum cost flow over all possible flows fij ∈ R as: However, as there is a minimization problem to solve, the computation time complexity is considerably high. Techniques providing efficient similarity search on large multimedia databases using the Earth Mover's Distance can, for instance, be found in [8,2,1]. Another approach for efficient indexing of the Earth Mover's Distance could be the metric space indexing.…”

Robust set reconciliation

“…Finally, there is a considerable amount of work on dealing with noise in data, particularly on data integration [14], similarity search [2], and joins [13]. But the problem of dealing with noise under the context of set reconciliation has not been studied before.…”

“…We note that the EMD can be more generally applied to other types of data, such as probability distributions and vectors [2]. Here we adopt its form when applied to measuring the distance between two point sets.…”

An efficient and effective similarity measure to enable data mining of petroglyphs

“…In soliciting feedback and advice for early previews of this work from various researchers in the data mining and image processing community, the feedback obtained was almost always of the form "Very nice, but have you considered using X", where X was Geometric Hashing (Wolfson and Rigoutsos 1997), Hausdorff Distance (Huttenlocher et al 1993), Chamfer Matching (Borgefors 1998), Shape Contexts (Belongie et al 2002), Fréchet Distance (Alt and Godau 1995), Skeleton Graphs (Bai and Latecki 2008), Zernike moments (Teague 1980), Earth Movers (Assent et al 2008), etc. While we have considered (and in some cases experimented with; see (Zhu 2009) these distance measures, space limitations prohibit a detailed review and discussion of the pros and cons of each of them.…”