EVALUATION OF EDGE DETECTION ALGORITHMS USING REAL IMAGES

Abstract

Edge detection is one of the earliest and most fundamental operations in most image understanding systems. Edge detection in an image significantly reduces the amount of data and filters out useless information, while preserving the important structural properties in an image. It is the most important module in the vision system. Literally thousands of edge detection algorithms have been described in literature. However, most new algorithms are described with no objective performance comparison to algorithms already exist. A variety of methods have been proposed for assessing the performance of edge detectors, some are theoretical while other need the ground truth for comparison. However, human rating experiments can be done in a more rigorous manner, to provide useful quantitative conclusions. The dissertation demonstrates the experimental strategy by comparing four well-known edge detectors: Canny, Rothwell, Laplacian and Sobel. It also answers the following questions: Is there a significant difference in edge detector outputs as perceived by humans? Do the edge detection results of an operator vary significantly with the choice of its parameters? For each detector, is it possible to choose a single set of optimal parameters for all the images without significantly affecting the edge output quality? Does an edge detector produce edges of the same quality for all images, or does the edge quality vary with the image? The algorithms are implemented in VC++ and five images of different variety are taken. The output to these images is taken with different parameter for Canny and Rothwell algorithm and best parameter for them are determined for each image. With best parameters all the output from four algorithms are compared.