摘要： 聚类是图像处理中的一个基本操作。对于分为K个簇的N个模式,且每个模式含有M个特征,该文在N个处理器的一维可重构流水线总线并行机上提出了一个时间复杂度为O(M×K)的聚类操作中一次迭代的并行算法。并对该并行算法进行了扩展,使得当处理器数分别为N×M和N×M×K时,一次迭代的时间复杂度分别为O(K)和O(1)。
关键词:
图像处理,
聚类,
平方误差技术,
并行算法
Abstract: Clustering is a fundamental operation in image processing.For N patterns and K centers,each with M features,this paper proposes a parallel algorithm which carries out one pass of the clustering in O(MK) time on a linear array with a reconfigurable pipelined bus system(LARPBS) with N processors.By increasing the number of processors,this proposed algorithm can be run in O(K) and O(1) time on LARPBS with NM and NMK processors,respectively.
中图分类号:
舒红霞;郑启龙;李春生;许胤龙;. 可重构流水线总线并行机上图像的聚类算法[J]. , 2005, 31(23): 166-168.
SHU Hongxia,ZHENG Qilong,LI Chunsheng,XU Yinlong. Algorithms for Image Clustering on Linear Array with Reconfigurable Pipelined Bus System[J]. , 2005, 31(23): 166-168.