We study the parallelization of the (record) linkage problem -i.e., to identify matching records between two collections of records, A and B. One of main idiosyncrasies of the linkage problem, compared to Database join, is the fact that once two records a in A and b in B are matched and merged to c, c needs to be compared to the rest of records in A and B again since it may incur new matching. This re-feeding stage of the linkage problem requires its solution to be iterative, and complicates the problem significantly. Toward this problem, we first discuss three plausible scenarios of inputs -when both collections are clean, only one is clean, and both are dirty. Then, we show that the intricate interplay between match and merge can exploit the characteristics of each scenario to achieve good parallelization. Our parallel algorithms achieve 6.55-7.49 times faster in speedup compared to sequential ones with 8 processors, and 11.15-18.56% improvement in efficiency compared to P-Swoosh.

Abstract. Finding near-duplicate images is a task often found in Multimedia Information Retrieval (MIR). Toward this effort, we propose a novel idea by bridging two seemingly unrelated fields -MIR and Biology. That is, we propose to use the popular gene sequence alignment algorithm in Biology, i.e., BLAST, in detecting near-duplicate images. Under the new idea, we study how various image features and gene sequence generation methods (using gene alphabets such as A, C, G, and T in DNA sequences) affect the accuracy and performance of detecting near-duplicate images. Our proposal, termed as BLASTed Image Linkage (BASIL), is empirically validated using various real data sets. This work can be viewed as the "first" step toward bridging MIR and Biology fields in the well-studied near-duplicate image detection problem.

No abstract

No abstract

No abstract