GPU-based 2D Optical flow using NVIDIA CUDA
Optical flow using variational methods which determine the unknown displacement field as a minimal solution of the energy functional.
In general, such energy-based formulations are composed of two parts: a data term which assumes constancy of specific image features, and a smoothness term which regularizes the spatial variation of the flow field.
Features
- Computational model guarantees:
- a unique solution (global optimal solution)
- stability of the algorithm (no critical dependence on parameters)
- High quality:
- dense flow results (one displacement for each pixel)
- allows large displacements
- different types of motion (translation, rotation, local elastic transformation)
- sub-pixel accuracy
- Robustness:
- under noise
- under varying illumination (brightness changes)
- with respect to artifacts
Examples
Figure: Left: Head of the feeding cockroach Periplaneta americana imaged by fast X-ray radiography and computed flow field, which captures the movements of the insect during chewing process. Right: Flow dynamics of liquid droplets in a fuel spray. Color coding: color represents direction and its brightness represents flow magnitude.
Model
- Brightness constancy data term [1]
- Gradient constancy data term [2]
- Data term based on higher-order derivatives [3]
- Robust modeling of data term [4]
- Flow-driven smoothness [3]
- Coarse-to-fine flow estimation [2]
- Intermediate flow median filtering [5]
References
- [1] B. Horn and B. Schunck. Determining optical flow. Artificial Intelligence, 17:185{203, 1981.
- [2] T. Brox, A. Bruhn, N. Papenberg, and J. Weickert. High accuracy optic flow estimation based on a theory for warping. In T. Pajdla and J. Matas, editors, Computer Vision, ECCV 2004, volume 3024 of Lecture Notes in Computer Science, pages 25-36. Springer, Berlin, 2004.
- [3] N. Papenberg, A. Bruhn, T. Brox, S. Didas, and J. Weickert. Highly accurate optic ow computation with theoretically justified warping. Int. J. Comput. Vision, 67(2):141-158, April 2006.
- [4] M. J. Black and P. Anandan. The robust estimation of multiple motions: Parametric and piecewise-smooth flow fields. Computer Vision and Image Understanding, 63(1):75- 104, 1996.
- [5] D. Sun, S. Roth, and M. J. Black. A quantitative analysis of current practices in optical flow estimation and the principles behind them. International Journal of Computer Vision, 106(2):115-137, 2014.