An improved Canny edge detection algorithm based on DCT

Abstract

The Canny edge detection algorithm is one of the most widely used in many computer vision and image analysis applications. However, it also tends to be sensitive to noise so that some the weak edge information can be easily lost when using the Gaussian filter to smooth images. In this paper, we propose a new method that replaces the traditional Gaussian filtering step in the Canny algorithm. In this method, the DCT coefficients are nonlinearly shrinked to achieve the denoising effect. Experiments show that the proposed methold produces better edge detection results at preserving more valid edge information and that it is more robust to noise than the traditional Canny edge detection algorithm.