基于NSCT及改进Canny的图像边缘检测

Edge detection of images based on NSCT and improved Canny

为进一步提取丰富的图像边缘信息,提出了一种基于非下采样Contourlet变换(Nonsubsampled Contourlet Transform,NSCT)及改进Canny的图像边缘检测方法。该方法是将图像进行NSCT多尺度分解,得到低频和高频子带。首先对低频子带使用改进Canny算子提取低频轮廓,再使用非线性函数对高频子带信息中各尺度各方向上的系数进行调整,实现边缘增强和噪声抑制,最后将NSCT域尺度内和尺度间的检测结果相融合来得到完整的边缘图像。实验结果表明,相比Sobel、Canny算子和现有的NSCT边缘检测方法,文中方法具有更好的边缘检测效果,边缘定位准确、完整、连续、细节丰富。

In order to extract more abundant image edge information,a method of image edge detection based on Nonsubsampled Contourlet Transform（ NSCT） and improved Canny was proposed.In this method,the image was decomposed in multiple scales by using NSCT to obtain the low frequency sub-band and high frequency sub-band.Firstly,the improved Canny operator was used to extract the low frequency contours for the low frequency sub-band,and then the nonlinear function was used to adjust the coefficients in the various directions of the high frequency sub-band to realize the edge enhancement and noise suppression.Finally,the complete edge image was obtained through the fusion of the measurement results of the same and different scales.Experimental results show that compared with Sobel,Canny and the existing NSCT edge detection method,the proposed method has better edge detection effect,and the edge location is accurate,complete,continuous and rich in details.