Vanishing point matching

Most vanishing point detection algorithms an based on the concept of clustering line intersections based on the density of line intersections in a local region. These methods are incapable of discriminating between true vanishing points and points which naturally arise as the mutual intersection of many lines in the image. We present a method of discriminating between non-vanishing points and vanishing points based on the identification of invariant properties of vanishing points in multiple images of a scene. Relaxation labelling is used to match the vanishing points of two images of a scene. The method is described and experimental results are provided which illustrate the successful performance of the algorithm