Download Stereo Code and Optical Flow Code

Abhijit S. Ogale

Staff Engineer, Computer Vision and Graphics

Home
Curriculum vitae
Publications
Research Teaching
Download Code

Download Stereo and Optical Flow code

The programs below compute symmetric dense stereo matching with occlusions, and dense optical flow with occlusions. Results are shown in the figure below. You are free to use these programs for any non-commercial research purposes. If you use these programs to get results for published work, please cite one of the papers below:

Shape and the stereo correspondence problem, A. S. Ogale and Y. Aloimonos, International Journal of Computer Vision, vol. 65, no. 3, 147-162, Dec 2005. (PDF).
A roadmap to the integration of early visual modules, A. S. Ogale and Y. Aloimonos, International Journal of Computer Vision: Special Issue on Early Cognitive Vision, vol. 72, no. 1, 9-25, Apr 2007. (PDF).

Running time for the above image pairs (top to bottom) on a 2.8 GHz Pentium PC:

Stereo (Tsukuba image pair): 1.2 sec
Stereo (SRI Trees pair): 0.5 sec
Optical flow (Flower garden sequence):4.6 sec

1. OpenVis3D - open source C++ library for dense matching

OpenVis3D is an open source C++ library for symmetric dense stereo with occlusions, dense optical flow with occlusions, and 3D egomotion estimation. It is available under the Apache License 2.0.

Download it here.

Credits: Thanks to Justin Domke for coding the 3D egomotion estimation. Thanks to Google, Inc. for providing us with an open-source grant for this work.

2. Matlab 7 code for dense stereo matching and dense optical flow (Win32/Linux)

These downloads contains some C routines precompiled to mex format using the Matlab compiler.

Example usage in Matlab:

Stereo matching code: (using example sawtooth images)
iLt = imread('sawleft.tif');
iRt = imread('sawright.tif');
shiftrange = [-19:-4];
[bestshiftsL, occlL, bestshiftsR, occlR] = stereoCorrespond(iLt, iRt, shiftrange);

Optical flow code: (using images from the flower-garden sequence)
i1 = imread('car1.png');
i2 = imread('car0.png');
shiftrangeX = [0:18];
shiftrangeY = [-3:0];
[x1,y1,o1,x2,y2,o2] = flow (i1, i2, shiftrangeX, shiftrangeY);

Note: Linux build environment: Matlab 7.1 R14 SP3, Debian Linux version 2.4.26-co-0.6.1, gcc version 3.3.3.

Other Code

For excellent egomotion estimation code, visit Justin Domke's page.

Download dissertation (thesis) template and LaTeX + LyX style files for the University of Maryland.

Copyright (c) 2003-2008 Abhijit Ogale
Dept. of Computer Science, University of Maryland, College Park

The software above is provided "as is", without warranty of any kind, express or implied, including but not limited to the warranties of merchantability, fitness for a particular purpose and noninfringement. In no event shall the authors or copyright holders be liable for any claim, damages or other liability, whether in an action of contract, tort or otherwise, arising from, out of or in connection with the software or the use or other dealings in the software.

Number of unique visitors since 12 Dec 2005:

Last updated: Feb 2008