An Adaptive Edge Detection Method Based On Gradient Histogram

Digital image plays an important role in people's life.Edge is the boundary between foreground and background in digital image,and it's an important feature of image.Edge detection has always been a basic and important research topic in many fields such as computer vision and image processing.There are many methods for edge detection has been developed,and Canny edge detection is the most perfect edge detection algorithm in theory.However,Canny algorithm has defects in steps of filtering and denoising,gradient calculation and double-threshold selection for edge connection,and the algorithm has low adaptability.For the defects of gaussian filtering in Canny algorithm,adaptive median filtering can be used instead of gaussian filtering to preserve more details of the image,improve the denoising ability of impulse noise and improve the adaptive ability of the algorithm.In the calculation of the gradient,the four-direction gradient calculation method is introduced to improve the anti-interference ability to noise,and at the same time,the image details can be effectively retained to reduce the probability of edge point omission.In the selection of double threshold,inspired by Otsu method,the double threshold Otsu method is deduced.On this basis,based on the gradient hisogram of the image,where the gradient has been processed through non-maximal suppression,the average number of non-zero pixels per gradient was obtained.Then the upper limit of the gradient probability of each gradient grade of the strong edge point is calculated according to the number.Finally,the upper limit of the high threshold in the double threshold is determined according to the probability.Compared with the Canny edge detection method that comes with MATLAB,the improved algorithm has better adaptive ability,stronger anti-interference ability to noise,richer details in the detection results and more accurate positioning of edge points.Moreover,the selected threshold values can be normalized and substituted into the Canny edge detection method of MATLAB,which also has better performance.