Edge detection using quantitative combination of multiple operators

Abstract

Although a number of diverse edge detection techniques can be found in many image processing publications, there is no single detection method that performs well in every possible image context. Information that could be missed by one detector may be captured by another. The purpose of this paper is to describe a new framework which allows us to quantitatively combine the methods of different edge detection operators in order to yield improved results for edge detection in an image. The so called receiver operating characteristics (ROC) analysis is employed in a novel fashion to form an optimum edge map that matches the outcomes of a preselected set of edge detectors. The results of applying the above ROC analysis technique are demonstrated and compared with individual edge detection methods.