Edge detection in noisy images with different edge types

Abstract

Edge Detection in an image is a process that produces edges of image objects, the purpose of which is to mark the parts that become detailed images to improve the details of blurry images, which occur due to the effects of the image acquisition process. Edge is defined as a change in the intensity of a large distance. Based on changes in intensity, there are three types of edges in digital images, namely, step edges, ramps edge, and noisy edge. On step edges where the intensity or gray value changes very fast The Gradient Method is able to detect better. On the ramp edge where the gray value slowly changes the Laplace method is able to detect better than the Gradient Method. On a noisy edge, the existence of noise in the image can bring up the other edges around the actual edge and can also shift the actual edge position. In this study, the authors conducted edge detection in the image by means of determining the right edge detection method to detect edges in noisy images. Noise is obtained by generating a Gaussian Noise in the image. The results showed that, without filtering the image, in noisy edge, the LoG operator was able to detect edges and reduce noise better than Canny. However, by selecting the right threshold that matches the σ (standard deviation), Canny also capable to provide good edge detection results.