摘要
图像分割的处理速度成为大规模图像数据处理的瓶颈。本文提出一种基于En FCM的图像聚类分割模型,直接对图像像素的灰度级进行聚类,能显著提高图像聚类分割的处理速度。为进一步提高处理速度,结合En FCM图像聚类分割模型特点,设计了三种并行优化策略——纯MPI并行方法、MPI+Open MP混合编程方法和CUDA并行架构方法 ,使其适合于大规模图像处理。实验结果表明,提出的三种并行优化策略都取得良好的加速效果。
The processing speed of image segmentation is the bottleneck for large image data processing. An image clustering segmentation model is proposed based on En FCM. It directly clusters grayscale of image pixel, clustering can significantly improve the image segmentation processing speed. In order to further improve the processing speed, three parallel design optimization strategies are designed by combining model features of En FCM image clustering segmentation, they are pure MPI parallel method,MPI ＋ Open MP hybrid programming and CUDA parallel architecture. These parallel methods are suitable for large-scale image processing. Experimental results show that the proposed parallel optimization strategies have achieved good acceleration effect.
出处
《微型机与应用》
2015年第15期55-58,共4页
Microcomputer & Its Applications
基金
茂名市科技计划项目(2014015)
