[1] M. Werlberger: Convex Approaches for High Performance Video Processing. PhD Thesis, Graz University of Technology, Institute for Computer Graphics and Vision. 2012.

[www]   [pdf]   [bibtex]

[2] M. Werlberger, T. Pock, and H. Bischof: Motion Estimation with Non-Local Total Variation Regularization, IEEE Conference on Computer Vision and Pattern Recognition (CVPR), San Francisco, CA, USA, June 2010.

[www]   [pdf]   [bibtex]

[3] M. Werlberger, W. Trobin, T. Pock, A. Wedel, D. Cremers, and H. Bischof: Anisotropic Huber-L1 Optical Fow. In Proceedings of the British Machine Vision Conference (BMVC), London, UK, September 2009.

[www]   [pdf]   [bibtex]


With the release of the FlowLib v3.0 we simplified some of the models' names. For documentation and parameters see the respective header files which are documented with doxygen (e.g. flowlib.h and parameters.h). For free usage we currently support Linux 64-bit, Win64 and Win32. In addition we compiled an application for simpler usage of the flow library (currently only supported for Linux 64-bit). Note that the library and also the application is only free for academic use. If you want to include our optical flow library for any other purpose contact Thomas Pock ( to talk about different license possibilities. There, also binaries for other platforms are available.

ImageUtilities - Bringing GPU and CPU closer together.


While rebuilding the optical flow library FlowLib we developed a new utility library that simplified the usage of images with CUDA. To provide this new library called ImageUtilities we decided to create a git repository and let everybody play around with this piece of code. The intention of doing so is of course to get some feedback on the functionality and motivate all of you to use the ImageUtilities and at the best include new features and functionality. If you have any questions and remarks on that feel free to contact me via e-mail (


You can download the code of the ImageUtilities library at Gitorious:   [GIT repository]