Techniques for Image Segmentation Based on Edge Detection
Weijie Zhou, Xiaoyu Du, Senhao Wang
Abstract
Image segmentation is a foundation in image processing, and edge detection directly affects image segmentation quality. However, there is no perfect and universal technique for image segmentation, and each method has its own feature. To compare the techniques, this paper conducted experiments of some edge detection fundamental techniques for image segmentation, which can be utilized to analyze the advantages and applicable scenes of various techniques. In this paper, experiments of edge detection based on methods such as ‘Sobel’, ‘Prewitt’, ‘Roberts’, ‘Canny’ method, and line hough transform are conducted using MATLAB. At the same time, this paper implements different thresholds on the Canny operator and compares the influence. As a result, we get different outputs according to these experiments. Canny performs better than the other three methods, and the low and high thresholds can influence the accuracy of the edge. Meanwhile, the hough transform can detect the line edge.