A Newly Improved Canny Algorithm of Image Edge Detection

A new adaptive Canny operator is proposed which aims at the effect of image edge areas when traditional Canny operator under Gaussian smooth and the shortcomings of threshold level need to be pre-set. Firstly, the adaptive median filtering is adopted to process image; Secondly, this paper adds the gradient template in the direction of 45 and 135 to calculate gradient amplitude; Finally, get the high and low threshold value of images by calculating gradient amplitude operation with the Otsu algorithm. In this paper, a series of experiments are designed and it verities the validity of the improved algorithm. The experiment results show that the improved Canny algorithm has a good anti-noise performance, and it can accurately detect the edge information. It reduces the existence of false edge and enhances the practicality of the algorithm. Introduction Image edge information is one of the most important information in an image, which can be described the target outline, the relative position within the target area, and other important information. Along with the development of computer vision and digital image processing technology, edge detection is widely used in all aspects of life, especially in the detection of the size of work piece, edge detection is an essential step. Most traditional edge detection methods take operation on the neighbor region pixels, and obtain the gradient with templates approximation, such as Prewitt, Robert, and Sobel, which are relatively simple and easy to implement, and it has a good real-time performance, but these operators are sensitive to noise, poor anti-interference performance, and their detection effect are not perfect in engineering application [1]. Compared with these edge detection algorithms, Canny operator is more accurate, and it has become the standard of evaluation of other edge detection method [2][3]. However, the traditional Canny algorithm adopts fixed spatial scale parameters for image smoothing. The high and low threshold of edge detection is lack of adaptability for different images. What’s more, the traditional Canny uses 2 2 × template to calculate the gradient, it is more sensitive to noise. Using iterative method to calculate the best threshold as the high-low threshold in literature [4], the improved algorithm has no obvious advantages to the noisy images and large scale of calculation, so it needs a long time and the results are poor. Literature[5] directly uses Canny Otsu algorithm of edge detection, which is easy detect of the false edges, not conducive to practical applications. Therefore, the paper proposes the improved Canny operator edge detection, the experiment results show that this given method has satisfactory results compared with the traditional Canny operator and the method in reference [5]. The Basic Principle and Analysis of the Traditional Canny Algorithm In 1986, John F. Canny proposed three criteria to judge image edge detection operator’ performance: SNR criterion, localization precision criterion and single edge response criterion [6]. The Traditional Canny Algorithm. The traditional canny image edge detection is divided into the following steps: the first step of traditional Canny algorithm is to smooth the original image by using a Gaussian filter, calculate the gradient magnitude and gradient direction; the second step is to adopt non-maximal inhibition after * Corresponding author, E-mail: yu_weibo@126.com 6th International Conference on Information Engineering for Mechanics and Materials (ICIMM 2016) © 2016. The authors Published by Atlantis Press 369 pretreatment of gradient image histogram, to refine those edges, and finally use fixed threshold to reduce the false edge points and to connect edges [7]. 1) image filtering Choose appropriate 1-d Gaussian function to smooth the image, according to the row and column respectively [8], the Gaussian filter function (as shown in (1)) to smooth image and remove noise.