Application of gradient-based edge detectors to determine vanishing points in monoscopic images: Comparative study

摘要

The detection of vanishing points in a monoscopic image is a first step to the extraction of 3D data. This article shows a partition of the image space in order to determine the type of perspective which is present, thereby allowing the determination of the vanishing point associated with each of the axes of the special reference system (X, Y, Z). Additionally, the Thales' second theorem allows us to determine the position of the vanishing points of the image and to automatize the process. An algorithm has been used with the data provided by the selected edge detector (Prewitt, Roberts, Sobel, Frei-Chen, Kirsch, Robinson 3 levels, Robinson 5 levels and Frei-Chen directional), which provides the location of the vanishing points contained on the image plane. The comparative study includes two variables: the number of vanishing points and the image resolution. The results show that in general the Prewitt's edge detector provides the best results, both positional and statistical. Increasing the image resolution improves the results, although the results obtained for a resolution of 640 x 480 pixels and another of 1024 x 768 pixels are very similar. (C) 2015 Elsevier B.V. All rights reserved.