Performance assessment of sobel and canny edge detectors.

Abstract

Abstract: This study evaluates the performance of Canny and Sobel Edge detectors using criteria based on Receiver Operating Characteristics (ROC) plots over a set of image patches synthetically generated, and explore how useful can be the Bayes risk associated to edge detector’s errors in order to set the threshold parameters which are responsible of the detector’s performance. The results indicate that Canny edge detector performs the best classification process of both detectors in this assessment. Furthermore, the Canny Edge detector turned out to be a better option than Sobel's because it has some additional stages of image processing beyond the point that Sobel edge detector acts.