By Topic

A parallel algorithm for graph matching and its MasPar implementation

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$31 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

5 Author(s)
Allen, R. ; Dept. of Comput. Sci. & Eng., Washington Univ., Seattle, WA, USA ; Yasuda, D. ; Tanimoto, S. ; Shapiro, L.
more authors

The authors have taken the forward-checking-graph-matching algorithm of L. G. Shapiro and R. M. Haralick (1981) and increased its speed and capacity, by developing a parallel approach and by improving the original algorithm. They discuss the parallel algorithm and its implementation on the MasPar single-instruction-stream/multiple-data-stream (SIMD) parallel processor, including job queues, job identification and control through the use of permutations, and load balancing. The combinatorial nature of graph-matching problems justifies more complex algorithms to eliminate as much of the search tree as possible. This work has increased the graph size for which solutions may be practically found, but the addition of a few more vertices quickly overwhelms the resources available. The parallel machine used is not powerful enough to justify its use in any real-time vision application; the problem size reached before outperforming the serial machine requires two to three minutes to solve. SIMD machines with 32-b processors and faster memory access would be much more competitive with current serial workstation technology. Converting this algorithm for a multiple-instruction-stream/multiple-data-stream (MIMD) parallel machine might produce results fast enough for real-time applications. However, one would expect this general mapping approach to object identification to always be outperformed by the more knowledge intensive constraint systems

Published in:

Computer Architectures for Machine Perception, 1993. Proceedings

Date of Conference:

15-17 Dec 1993