文摘
The wavelet transform which is very famous provides a very effective framework for edge detection has a limited ability in handling the directional information of images. In this paper, we propose a method based on the nonsubsampled contourlet transform which is a multiscale and multidirectional transform with a perfect ability to detect distributed discontinuities such as edges and denoise. Nonsubsampled (NS) contourlet transform is theoretically optimal in representing images with edges especially has the ability to fully capture directional and other geometrical features, which is different from the traditional wavelet transform. Numerical experiments demonstrate that the edge detector proposed in this paper based on the nonsubsampled contourlet transform is highly effective in detecting edges, and outperforms traditional method based on wavelets as well as other famous methods.