首页 | 官方网站   微博 | 高级检索  
     

基于五阶CNN的图像边检测算法研究
引用本文:李国东,王雪,赵国敏.基于五阶CNN的图像边检测算法研究[J].安徽大学学报(自然科学版),2015(3):15-21.
作者姓名:李国东  王雪  赵国敏
作者单位:新疆财经大学应用数学学院,新疆乌鲁木齐,830012
基金项目:国家自然科学基金资助项目(11461063);国家社会科学基金资助项目(14B T J021);国家教育部人文社会科学基金资助项目(13YJAZH040);新疆维吾尔自治区高校科研计划项目(XJEDU2013I26);新疆财经大学研究生科研创新项目(2014045);新疆财经大学硕士研究生科研创新重点科研项目
摘    要:边缘是图像的最基本的特征之一,边缘提取是图像分析中非常重要的步骤,而细胞神经网络是边缘检测中很有效的一种方法.作者基于细胞神经网络(cellular neural network,简称CNN),研究了5阶CNN模板对图像边缘检测的过程,阐述了算法实现过程中的关键步骤,并且证明了算法的稳定性.对图像分别采用基于5阶、3阶CNN算法和经典算子(Prewit、Canny、Sobel等)进行边缘提取,定性分析比较了几类算法在性能上的优劣,定量比较了检测结果的准确性.实验结果表明,基于5阶CNN模板算法的边缘检测结果更加显著,且在硬件实现上能够高速并行计算,实现图像实时处理.

关 键 词:CNN  边缘检测  5阶模板  品质因子P

The algorithm for detection edge of image based on the 5 dimension CNN
LI Guo-dong , WANG Xue , ZHAO Guo-min.The algorithm for detection edge of image based on the 5 dimension CNN[J].Journal of Anhui University(Natural Sciences),2015(3):15-21.
Authors:LI Guo-dong  WANG Xue  ZHAO Guo-min
Abstract:Edge is one of the basic characteristics of image ,edge detection is a very important step in image analysis ,and cellular neural network as a method is very effective in edge detection .This article was based on cellular neural networks (cellular neural network , CNN) ,studied the 5 dimenson CNN template about the process of image edge detection , expounded the key steps in the process of algorithm realization ,and proved the stability of the algorithm .With image respectively based on 5 dimenson CNN algorithm ,3 dimenson CNN algorithm and classical algorithm (Prewit ,Canny ,Sobel ,etc .) for edge detection ,the article analysed and compared the pros and cons of several kinds of algorithm on the performance , the accuracy of the quantitative comparison of the test results . T he experimental results showed that the CNN template algorithm of edge detection based on 5 dimenson results was more effective , with high‐speed parallel computing in hardware implementation ,could achieve real‐time image processing .
Keywords:cellular neural networks (CNN)  edge detection  5 dimenson template  quality factor P
本文献已被 CNKI 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司    京ICP备09084417号-23

京公网安备 11010802026262号