A novel multi-scale and multi-expert edge detector based on common vector approach

In this paper, an information fusion algorithm – MSME-CVAED – is proposed for multi-scale and multi-expert edge detection. Well-known operators, called experts, have been applied to distinct scales derived by smoothing the gray image with Gaussian functions having different variance values. Common characteristics of edge points are processed to merge the information obtained from each scale, based on the concept of common vector approach. Once a single gradient map obtained, a smart non-maximum suppression operation is carried out to obtain a binary edge map. Later, an edge segment validation process is introduced based on Helmholtz principle, a common method in which edge segment validation is carried out with the “a contrario” approach using the number of false alarms concept. Experiments on popular datasets of ROC curves and RUG show that the proposed method achieves superior results in terms of F-measure and Mathews correlation coefficient score, compared with some recently published edge detectors.