Robust dense optical flow.

For non-commercial uses, please email me and I will send you the code free of charge.

For commercial uses, please contact me regarding licensing.

Password protected tar file.

Note: if you use Explorer to download this and it gets saved as denseFlow.tar.tar just rename the file to denseFlow.tar.gz and then unzip it. I don't know why this happens with Explorer.


This tar file contains experimental C-code, and examples, for computing dense optical flow fields robustly.

This code computes a dense optical flow field using a robust estimation technique. The basic method uses a coarse-to-fine gradient-based flow computation much like the Horn and Schunck algorithm. The difference is that the problem is formulated robustly to allow violations of the brightness constancy and spatial coherence assumptions.

This is a very basic version of the code. My hope is that it helps illustrate the main ideas involved with robust regularization. In addition to being illustrative, it actually computes decent flow fields.


Black, M. J. and Anandan, P., The robust estimation of multiple motions: Parametric and piecewise-smooth flow fields, Computer Vision and Image Understanding, CVIU, 63(1), pp. 75-104, Jan. 1996. (abstract)

Black, M. J. and Anandan, P., A framework for the robust estimation of optical flow, Fourth International Conf. on Computer Vision, ICCV-93, Berlin, Germany, May, 1993, pp. 231-236. (postscript)