Morphological operators that considerably improve the generalised hough transform

Abstract Hough Transform has difficulty detecting short lines. One way to improve the results is to oversample images in order to make those short lines longer. Unfortunately, this method is computationally expensive. In this article, three new binary morphological operators are introduced that can stretch binary edge images. The first operator thins the output of the edge detector so that the second one can stretch it. The Stretch operator does not alter the coordinates of the end points of a line (other than multiply by 3) nor make it thicker. The third one, the smooth operator, is designed for smoothing the jaggedness of the lines stretched by the stretch operator. These operators are a computationally efficient way of improving the results of the Hough Transform for lines, circles and other shapes. All of the operators are implemented using the OpenCV library and the source code is publicly shared on GitHub.