分散メモリシステムにおける大規模距離画像の並列同時位置合わせ手法

This paper describes a method for parallel alignment of multiple range images. Since it is difficult to align a large number of range images simultaneously, we developed a parallel method to accelerate and reduce the memory requirement of the process. Although a general simultaneous alignment algorithm searches correspondences for all pairs of all range images, by rejecting redundant dependencies, our method makes it possible to accelerate computation time and reduce the amount of memory used on each node. The correspondence search is performed independently for each pair of range images. Accordingly, the computations between the pairs are preformed in parallel on multiple processors. All relations between range images are described as a pair node hyper-graph. Then, an optimal pair assignment is computed by partitioning the graph properly. The method was tested on a 16 processor PC cluster, where it demonstrated the high extendibility and the performance improvement in time and memory.