Parallel traversal of signature trees for fast CBIR
Aaron Davidson, John Anvik, Mário A. Nascimento · 2001
This paper presents a technique for searching a memory-resident signature tree (S-tree) in a parallel fashion. Our implementation uses a very compact binary signature for the images' global colour histogram thus allowing the use of main-memory to store and search a reasonably sized image main-memory to store and search a reasonably sized image collection. Our results show that our parallel search of an S-tree is able to achieve good speedups; for instance, a 50,000 signature S-tree, with a branching factor of 8, searched using 4 processors is 3 times faster than using a single processor.