Heterogeneous Computing for Real-Time Stereo Matching

Stereo matching is used in many computer vision applications such as 3D reconstruction, robot navigation, robotic surgery, 3-D video surveillance, and tracking object in 3D space. Real time stereo matching is difficult due to the heavy computation required for matching algorithms. In this paper a CPU/GPU heterogeneous computing platform is used to accelerate the processing and run the system in real time. The availability of GPUs (Graphics Processing Units) with hundreds of parallel processing cores, increases the speed of matching algorithms. In this work a combination between feature based and area based matching techniques is used. Feature based matching is fast while area based is robust and the used technique takes the advantages of the both. The stereo matching algorithms run over GeForce GPU and a real time processing is achieved with speed up to 15 frames/sec.

[1]  Aly A. Farag,et al.  Cooperative stereo: combining edge- and area-based stereo , 1999, 1999 IEEE Aerospace Conference. Proceedings (Cat. No.99TH8403).

[2]  Keqin Li,et al.  Parallel Algorithms for Approximate String Matching with k Mismatches on CUDA , 2012, 2012 IEEE 26th International Parallel and Distributed Processing Symposium Workshops & PhD Forum.

[3]  Eric Psota,et al.  Real-Time Stereo Matching on CUDA Using an Iterative Refinement Method for Adaptive Support-Weight Correspondences , 2013, IEEE Transactions on Circuits and Systems for Video Technology.

[4]  Yingyun Yang,et al.  A new stereo matching algorithm based on adaptive window , 2012, 2012 International Conference on Systems and Informatics (ICSAI2012).

[5]  Soon-Yong Park,et al.  Fast Stereo Matching of Feature Links , 2011, 2011 International Conference on 3D Imaging, Modeling, Processing, Visualization and Transmission.

[6]  Sumam David,et al.  A novel adaptive support window based stereo matching algorithm for 3D reconstruction from 2D images , 2011, 2011 11th International Conference on ITS Telecommunications.