互联网 qkzz.net
全刊杂志网:首页 > 女性 > 文章正文
刊社推荐

基于CNN的分块自适应彩色图像边缘检测的研究姜庆玲 刘万军 张 闯


   (1.辽宁工程技术学院 a. 电子与信息工程学院; b. 软件学院,辽宁 葫芦岛125105; 2.铁岭师范高等专科学校 理工学院, 辽宁 铁岭 112300; 3.北京科技大学 信息工程学院, 北京100083)

  

  摘 要:利用细胞神经网络(CNN)模型对彩色图像边缘检测时,首先要解决彩色空间的选择以及颜色距离的计算问题,其次网络参数的选择也是一个重要问题。为了达到在确保边缘检测准确的同时有效抑制噪声的目的,对整幅图像进行分块自适应检测,采用熵来度量图像的各个子区域的不同性质,然后根据该区域的性质选择一组合适的网络参数,对提取该区域图像边缘的CNN 模板进行了理论分析和鲁棒性研究,提出一个设计符合相应功能要求的CNN 鲁棒性定理,它为设计相应的 CNN 模板参数提供了解析判据。仿真实验表明,该算法具有较好的健壮性。

  关键词:细胞神经网络; 边缘检测; 熵; 人类视觉系统; 鲁棒性

  中图分类号:TP391 文献标志码:A

   文章编号:10013695(2009)03113104

  

  Color edge detection based on CNN’s block adaptive

  

  JIANG Qingling1a,2, LIU Wanjun1b, ZHANG Chuang3

  

  (1.a. Shool of Electronics & Information Engineering, b.School of Software Engineering, Liaoning Technical University, Huludao Liaoning 125105,China; 2.School of Polytechnic Engineering,Tieling Teachers College,Tieling Liaoning 112300, China; 3.School of Information Engineering, University of Science & Technology Beijing, Beijing 100083,China)

  

  Abstract:The researches and applications of image processing based on the CNN have made great progresses. Edge detection is one of the basic preprocessing methods in digital image processing and computer vision. This paper proposed a color edge detection scheme based on CNN templates, and then studied and analyzed robustness CNN templates. First of all to solve the problem of the choice of color space and the problem of the calculation of distance color, followed by the choice of network parameters was also an important issue. In order to achieve accurate edge detection in ensuring the effective suppression of the noise, on the whole image block adaptive testing, used entropy to measure the various images of the different nature of the region, according to the nature of the region chose a suitable network parameters.It presented a theorem for designing the robustness CNN template parameters. Experiment results indicate that the approach has extensive application scope.

......
很抱歉,暂无全文,若需要阅读全文或喜欢本刊物请联系《计算机应用研究》杂志社购买。
欢迎作者提供全文,请点击编辑
分享:
 

了解更多资讯,请关注“木兰百花园”
分享:
 
精彩图文


关键字
支持中国杂志产业发展,请购买、订阅纸质杂志,欢迎杂志社提供过刊、样刊及电子版。
关于我们 | 网站声明 | 刊社管理 | 网站地图 | 联系方式 | 中图分类法 | RSS 2.0订阅 | IP查询
全刊杂志赏析网 2017