COMPARATIVE STUDY OF EDGE DETECTION ALGORITHMS USING REAL IMAGES

Abstract

Digital image processing is the manipulation of images by a computer. Feature extraction on the other hand is a process wherein specific features in an image like edges, boundaries, shape etc. are extracted for identification either visually or by a computer. This work mainly concentrates on edge detection. A comparative study of five edge detection algorithms namely Sobel, Prewitt, Line detection, Laplacian and Line Edge Map (LEM) is presented. The basic measure of performance is visual rating, which . indicates the perceived quality of the edges for identifying an object. The process of evaluating edge detection algorithms with this performance measure requires the collection of a set of gray-scale real images, applying the edge detection algorithms on each image and conducting visual evaluation of the resultant images so obtained. Assessing the performance of edge detection algorithms is difficult because the performance depends on several factors such as the algorithm itself, the type of images used to measure the performance of the algorithm and the human judgement and estimate. To ensure a good comparative study, the real images of varying scenes are used in the evaluation of the edge detection algorithms. A set of images was collected that contain objects that people can readily recognize and name. This was done because the object recognition task to be performed with the edge images would not be meaningful if people could not recognize the .object in an image. The results of the five edge detection algorithms are also compared with commercially available software Adobe PhotoShop 7.0. The results and their interpretation are discussed in chapter 5.