The focus of this project is the development of a cooperative stereo vision algorithm for obtaining disparity maps and explicitly detecting occlusions. To produce smooth and detailed disparity maps, our algorithm utilizes two assumptions: uniqueness and continuity. That is, the disparity maps have a unique value per pixel and are continuous almost everywhere.
Our cooperative algorithm enforces these assumptions within a 3D array of match values in disparity space (row, column, disparity.) An iterative algorithm refines the match values by diffusing support among neighboring values and inhibiting others along similar lines of sight.
Software
Source Code (C++) now available 2/6/03 - here
If you plan on using our source code in your research please email clz@cs.cmu.edu stating how you plan on using the code. The code may only be used for research and non-commercial purposes. Within any publications please reference:
A Cooperative Algorithm
for Stereo Matching and Occlusion Detection
C. L. Zitnick and T. Kanade
Proc. IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 22,
no. 7, July 2000.
Software Update 10/24/01 - Bug fixed, no longer crashes after computing disparity map with Windows 2000.
An executable is available for running the stereo algorithm. It runs on Windows NT 4.0 with 24 or 32 bit color. Directions for use along with the program are contained in ZK Stereo.zip. This software may only be used for research and educational purposes, please read README.txt.
For sample stereo image pairs please visit the CMU
image database.
The results on this page were produced as described in Zitnick and Kanade, 1999. Any questions regarding the algorithm should be sent to clz@ri.cmu.edu.
Head scene, University of Tsukuba
Pentagon Scene
Coal Scene
Shrub Scene
Town Scene
C. Lawrence Zitnick - PhD Student, RI
Takeo Kanade - Director/Robotics, U. A. Helen Whitaker University Professor, RI/CS
A Cooperative Algorithm
for Stereo Matching and Occlusion Detection
C. L. Zitnick and T. Kanade
Proc. IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 22,
no. 7, July 2000.
A Cooperative Algorithm
for Stereo Matching and Occlusion Detection
C. L. Zitnick and T. Kanade
tech. report CMU-RI-TR-99-35, Robotics Institute, Carnegie Mellon University,
October, 1999.
A Volumetric Iterative
Approach to Stereo Matching and Occlusion Detection
C. L. Zitnick and T. Kanade
tech. report CMU-RI-TR-98-30, Robotics Institute, Carnegie Mellon University,
December, 1998.
CMU's Vision and Autonomous
Systems Center
|
|
