脉冲耦合神经网络的并行加速优化算法研究  

The Study of Parallel Programming Optimization of Pulse Coupled Neural Network

作　　者：钟瑾[1] 施俊[1] 常谦[1] 

机构地区：[1]上海大学通信与信息工程学院,上海200072

出　　处：《生物医学工程学进展》2010年第1期7-11,共5页Progress in Biomedical Engineering

基　　金：国家自然科学基金资助项目(60701021);上海市教育委员会科研创新项目(09YZ15);上海市教委重点学科建设项目(J50104)

摘　　要：并行编程技术可以有效提高算法的执行效率。文中分别利用CPU的单指令多数据流扩展指令集(Streaming SIMD Extensions,SSE)技术和多核并行编程技术,对脉冲耦合神经网络(Pulse Coupled Neural Network,PCNN)分割算法进行并行编程优化,以减少算法的运行时间。实验结果表明,SSE技术以及多核并行编程技术大大加快了PCNN分割算法的运行速度,有效提高了算法的执行效率,在一定程度上解决了该方法计算量大、耗时多的问题,具有应用于医学图像处理的潜在价值。The parallel programming technique can improve the algorithm＇s running efficiency.In this study,the Streaming SIMD Extensions（SSE）instructions in CPU and multi-core parallel programming technique were respectively used to optimize the pulse coupled neural network（PCNN）based segmentation algorithm so as to reduce the execution time.The experimental results indicated that SSE and multi-core parallel programming technique could speed up the PCNN-based segmentation algorithm,improve the running efficiency,and solve the problems of large computation and long time consume for PCNN at a certain extent.The results suggest that parallel programming has the potential to be used in medical image processing.

关 键 词：脉冲耦合神经网络 单指令多数据流扩展指令集(SSE) 多核 并行编程 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]