This is the MATLAB software package for horizon detection on a single image paper Xu et al. CVPR 2013. Use the following citation:

Xu, Yiliang, Sangmin Oh, and Anthony Hoogs. 
"A minimum error vanishing point detection approach for uncalibrated monocular images of man-made environments." 
In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 1376-1383. 2013.

BibTex

@InProceedings{Xu_2013_CVPR,
	author = {Xu, Yiliang and Oh, Sangmin and Hoogs, Anthony},
	title = {A Minimum Error Vanishing Point Detection Approach for Uncalibrated Monocular Images of Man-Made Environments},
	booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
	month = {June},
	year = {2013}
}

Copyright 2014 by Kitware, Inc. All Rights Reserved. Please refer to LICENSE.TXT for licensing information, or contact General Counsel, Kitware, Inc., 28 Corporate Drive, Clifton Park, NY 12065

Software Structure

horizon_detection.m contains the main function. - extract_lsd.m: calls Line Segment Detector (LSD) C++ executable "LSD_VC.exe" to compute line segments in the input image. LSD_VC.exe read in an image and output a text file containing all detected line segments information. The Matlab script reads in this text file. This module can be substituted by other line segment detection algorithm. - vp_probability_EM.m: compute vanishing points given line segments that are detected. - horizon.m: compute horizon given vanishing points that are detected.

Dependency

- Line Segment Detector (LSD) C++ executable "LSD_VC.exe" is included in this package. The source files can be found at http://www.ipol.im/pub/art/2012/gjmr-lsd/.
- MATLAB basic package
- MATLAB image processing toolbox (only required for visualizing results).