DGW-canny: An improvised version of Canny edge detector

Edge detection is a field which has intrigued programmers since early 1970s. Since then, a good number of edge detection techniques have come up but the technique used by Canny [1] is very widely used. However, it has been observed that the results are not that efficient while dealing with alpha-numeric sections or geometrical figures or fine grain region in an image. To mitigate these limitations, we propose an improvised edge detection technique. The technique uses a Laplacian of Gaussian gradient with a new approach towards the thresholding section. The DGW-Canny framework has been experimented on a data set of images categorized on the basis of vectors proposed by us. The encouraging experimental results show that the DGW-Canny edge detector renders much better results compared to those of Canny. We have also performed experiments by varying parameters like nature and man-made component in an image. It is observed that varying the latter parameter adversely affects the edges of the image.