基于CUDA的图像快速并行细化算法的研究与实现  

Research and implementation of image thinning algorithm on CUDA platform

作　　者：金汉均[1] 梅洪洋 

机构地区：[1]华中师范大学计算机学院,武汉430079

出　　处：《电子测量技术》2014年第8期75-79,共5页Electronic Measurement Technology

基　　金：湖北省科技攻关(011EJB010);湖北省科技支撑计划(2013BAA104)项目

摘　　要：在图像处理和模式识别中,通过细化算法抽取图像的骨架是一项极有意义的工作。由于大多数传统细化算法都需要按照一定顺序对每个像素进行逐点扫描,执行效率不高,另外图像细化算法中数据之间的关联性不大,将传统的快速并行细化算法移植到CUDA(compute unified device architecture)平台上,运用CUDA的并行处理优势对图像进行细化处理。实验表明,该方法在保证细化处理效果的前提下,还能将运算效率提高达到幂级数级。In the image processing or the pattern recognition, extracting the image skeleton through thinning algorithm is a very meaningful work. As most of the traditional algorithms need to scan the pixel point by point with its poor efficiency,and for another way, the connection between image data is weak, so this paper transplant the thinning algorithm to CUDA platform and run with its parallel computing. Experiments show that the operation efficiency is improved under the precondition of the high precision of the image.

关 键 词：细化算法 CUDA平台 并行处理 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]