X. Fu, K.J. Liszka, and B. Xie (USA)
parallel algorithms, wavelet packet transforms, loadbalancing, cluster computing
We present an optimal parallel algorithm for a wavelet packet transform tree decomposition. Fast computational speed is achieved by dispatching tasks to multi-processors efficiently on a cluster of computers. Significant speedup is achieved through an optimal dispatch pattern based on binary format features, with good load balancing and minimal communication requirements. Implementation of the algorithm and complexity analysis show that this algorithm is highly scalable and is cost optimal as the problem size increases. Experimental results are obtained from experiments for locating pinpoint holes in aluminum plates.
Important Links:
Go Back
