关键词: 非下采样Contourlet变换, 最大类间方差, 自适应方法, Canny算子

Abstract: In order to fix the inconvenience that in traditional Canny operator，human intervention is required to select the Gaussian filter parameter and the high and low threshold，this paper proposes a self-adaptive edge detection method. Instead of applying Gaussian filter，this method uses self-adaptive thresholding algorithm，which is based on Nonsubsampled Contourlet Transform（NSCT） algorithm，to denoise an image and keep its edge information.Then the Otsu method sets high and low thresholds automatically，which are applied to edge detection and connection of the image generated by non-maximal suppression.The theoretical and experimental results show that this method，compared with the traditional Canny operator，not only filters out image noises and keeps detail edges efficiently，but also improves its automation.

Key words: Nonsubsampled Contourlet Transform（NSCT）, Otsu, self-adaptive method, Canny operator