Establishing Straight Line Correspondence
Manolis Lourakis · 1997
Straight line segments are common in images of man-made environments. Due to their mathematical simplicity and the ease with which they can be extracted from images, they have been widely used in vision applications. This paper presents a method for extracting and corresponding line segments between two views. Line segments are extracted by detecting edges and then fitting straight lines to the resulting edge points. For each of the line segments extracted from one image, a set of potential matches in the other image is formed. The strength of each potential match is measured by a matching function based on a few line attributes. A relaxation labeling technique that imposes global constraints on line matches is then used to disambiguate line matches. Results from the application of the method on real images are also presented.