摘要
针对现有的边缘提取方法存在着无法同时满足抑噪能力强和定位精度高的问题,提出了多尺度的Her-mite变换边缘检测法.通过分析可以证明Hermite法是一种建立在滤波基础上的边缘提取方法,而且由于Her-mite变换提取出的是图像局部最优方向下的边缘检测,因此其在提取边缘时的抑噪能力要比一般方法高.此外,由于多尺度Hermite变换结合了大尺度的抑噪能力和小尺度的定位能力,因此这种方法能非常有效地提取出边缘并保持了较高的定位精度.通过与Sobel法的边缘检测结果进行对比,可以发现多尺度Hermite变换法在进行低信噪比条件下的边缘检测时具有较好的抑噪和定位能力.
The existent methods of edge detection can＇t satisfied these roles simultaneously that are suppressing noise and precise orientation. So the existent methods are ineffective and need to be improved. In this article, a multiscale Hermite transform method （MHT） has been put forward to solve these problems. By analysing Hermite transform, it can be proved that Hermite transform is based on Gaussian filtering and represents the optimized direction of local area, so MHT method has better expression than other methods in noise suppression. Furthermore, MHT method fused the character of big scale and small scale. These characters are that noise is well suppressed and the orientation of edge is precise. At last, comparisons between the MHT method and the often used Sobel algorithms are made. The results indicate that the good noise suppression and precise orientation can be got, when this MHT method is used to detect edge of Iow-SNR image.
出处
《武汉大学学报（工学版）》
CAS
CSCD
北大核心
2006年第2期47-50,54,共5页
Engineering Journal of Wuhan University
基金
航空基础科学基金(04I53067)
高等学校博士学科点专项科研基金(20020699014)资助项目
