Vanishing Point Detection and Camera Calibration

Bo Li (prclibo@gmail.com), Kun Peng, Xianghua Ying, Hongbin Zha

For images taken in man-made scenes, vanishing points and focal length of camera play important roles in scene understanding. In this paper, we present a novel method to quickly, accurately and simultaneously estimate three orthogonal vanishing points (TOVPs) and focal length from single images. Our method is based on the following important observations: If we establish a polar coordinate system on the image plane whose origin is at the image center, angle coordinates of vanishing points can be robustly estimated by seeking peaks in a histogram. From the detected angle coordinates, altitudes of a triangle formed by TOVPs are determined. Novel constraints on both vanishing points and focal length could be obtained from the three altitudes. By using the constraints, radial coordinates of TOVPs and focal length can be estimated simultaneously. Our method decomposes a 2D Hough parameter space into two cascaded 1D Hough parameter spaces, which makes our method much faster and more robust than previous methods without losing accuracy. Enormous experiments on real images have been done to test feasibility and correctness of our method.
 
In recent research, we extended the above algorithm to a fast vanishing point detection algorithm based on 1D histogram. This new algorithm does not requrire that the TOVPs are all far away from the image center. We divide an image plane into interior and exterior regions. Typical 2D histograms corresponding to these two regions could be simplified into some 1D histogram. We develop some constraints on the simplification based on ``Manhattan world'' assumption so that it will not lose useful information. We test our alogrithm and some commonly used vanishing point detection on public database YorkUrbanDB and our own implemented database. Our algorithm shows significant performance improvement
 
Publication
 
Bo Li, Kun Peng, Xianghua Ying, Hongbin Zha. Simultaneous Vanishing Point Detection and Camera Calibration from Single Images. Accepted for oral presentation by the 6th International Symposium on Visual Computing, 2010. [PDF]
Bo Li, Kun Peng, Xianghua Ying, Hongbin Zha. Vanishing Point Detection Using Cascaded 1D Hough Transform from Single Images, In progress.
 
Software
 
Vanishing point detection algorithms of our algorithm and some previous research could be downloaded here: VPD.zip, which inludes:
ISVCDemo A simple MATLAB code demo for our ISVC10 paper.
Bo Li, Kun Peng, Xianghua Ying, Hongbin Zha, Simultaneous Vanishing Point Detection and Camera Calibration from Single Images, Proc. International Symposium on Visual Computing 2010, Part II, LNCS 6454, pp. 151--160.
IERDemo MATLAB code of our complete vanishing point detection algorithm.
EMDemo
MATLAB code of vanishing point detection algorithm proposed by
J. Kosecka and W. Zhang. Video compass. Proc. European Conference on Computer Vision, pp. 657-673, 2002.
GSDemo
MATLAB code of vanishing point detection algorithm proposed by
S. T. Barnard. Interpreting perspective images. Artificial Intelligence, vol. 21, pp. 435-462, 1983.

Another vanishing point detection algorithm using the "J-Linkage Algorithm" could be found from http://www-etud.iro.umontreal.ca/~tardifj/

Database

Our own implemented groundtruth database PKU Campus Database for vanishing point detection could be downloaded here: PKUCampusDB.zip, which includes 200 photos of man made scene and corresponding vanishing points groundtruth. Note that all photos in our database have three orthogonal vanishing points.

Another public groundtruth databse we used to test vanishing point detection algorithms is the York Urban Database, which could be downloaded here: http://www.elderlab.yorku.ca/YorkUrbanDB/

กก

กก