| Optical flow evaluation results |
Statistics:
Average
SD
R2.5
R5.0
R10.0
A50
A75
A95
Error type: angle endpoint interpolation normalized interpolation |
|
Average angle error |
avg. |
Army (Hidden texture) GT im0 im1 |
Mequon (Hidden texture) GT im0 im1 |
Schefflera (Hidden texture) GT im0 im1 |
Wooden (Hidden texture) GT im0 im1 |
Grove (Synthetic) GT im0 im1 |
Urban (Synthetic) GT im0 im1 |
Yosemite (Synthetic) GT im0 im1 |
Teddy (Stereo) GT im0 im1 |
||||||||||||||||
| rank | all | disc | untext | all | disc | untext | all | disc | untext | all | disc | untext | all | disc | untext | all | disc | untext | all | disc | untext | all | disc | untext | |
| Classic+Area [31] | 5.4 | 3.36 2 | 10.0 4 | 2.85 3 | 2.93 6 | 11.1 8 | 2.28 6 | 6.84 14 | 16.4 15 | 3.06 9 | 2.98 2 | 16.9 6 | 1.42 1 | 3.06 1 | 4.06 2 | 2.36 1 | 3.59 2 | 10.9 1 | 2.86 3 | 2.76 11 | 3.54 6 | 3.10 20 | 1.90 1 | 4.09 1 | 2.85 5 |
| NL-TV-NCC [29] | 5.8 | 3.89 7 | 9.16 1 | 2.98 4 | 2.87 5 | 9.69 3 | 1.99 2 | 4.44 5 | 11.6 5 | 1.76 1 | 2.64 1 | 11.8 1 | 1.48 2 | 3.49 7 | 4.60 12 | 2.47 3 | 4.67 10 | 13.5 4 | 4.26 11 | 2.83 13 | 4.57 17 | 2.84 16 | 2.62 3 | 6.00 3 | 2.25 2 |
| Complementary OF [24] | 6.4 | 4.44 13 | 11.2 10 | 4.04 14 | 2.51 2 | 9.77 4 | 1.74 1 | 3.93 4 | 10.6 4 | 2.04 2 | 3.87 9 | 18.8 8 | 2.19 8 | 3.17 2 | 4.00 1 | 2.92 7 | 4.64 9 | 13.8 5 | 3.64 8 | 2.17 5 | 3.36 3 | 2.51 13 | 3.08 5 | 7.04 4 | 3.65 13 |
| Adaptive [23] | 6.8 | 3.29 1 | 9.43 2 | 2.28 1 | 3.10 9 | 11.4 10 | 2.46 9 | 6.58 12 | 15.7 12 | 2.52 7 | 3.14 3 | 15.6 3 | 1.56 3 | 3.67 12 | 4.46 7 | 3.48 11 | 3.32 1 | 13.0 3 | 2.38 1 | 2.76 11 | 4.39 15 | 1.93 9 | 3.58 8 | 8.18 7 | 2.88 6 |
| Aniso. Huber-L1 [25] | 8.5 | 3.71 4 | 10.1 5 | 3.08 5 | 4.36 16 | 13.0 12 | 3.77 16 | 6.92 15 | 15.3 10 | 3.60 17 | 3.54 5 | 15.9 4 | 2.04 6 | 3.38 4 | 4.45 6 | 2.47 3 | 3.88 4 | 12.9 2 | 2.74 2 | 3.37 18 | 4.36 14 | 2.85 17 | 3.16 6 | 7.52 5 | 2.90 7 |
| DPOF [18] | 9.1 | 5.12 17 | 12.9 17 | 3.49 9 | 3.07 8 | 10.3 5 | 2.44 8 | 3.09 1 | 7.47 2 | 2.43 6 | 3.42 4 | 12.9 2 | 2.41 13 | 3.55 9 | 4.56 11 | 3.35 9 | 4.69 11 | 14.2 6 | 5.14 12 | 3.59 20 | 4.67 19 | 3.83 24 | 2.00 2 | 4.93 2 | 1.65 1 |
| Spatially variant [19] | 9.5 | 3.73 5 | 10.2 7 | 3.33 6 | 3.02 7 | 11.0 7 | 2.67 10 | 5.36 8 | 13.8 9 | 2.35 3 | 3.67 6 | 19.3 10 | 1.84 5 | 3.81 16 | 4.81 20 | 3.69 14 | 4.48 8 | 16.0 12 | 3.90 9 | 2.11 3 | 3.26 2 | 2.12 11 | 4.66 17 | 9.41 15 | 4.35 17 |
| TV-L1-improved [17] | 10.1 | 3.36 2 | 9.63 3 | 2.62 2 | 2.82 4 | 10.7 6 | 2.23 3 | 6.50 11 | 15.8 13 | 2.73 8 | 3.80 8 | 21.3 17 | 1.76 4 | 3.34 3 | 4.38 5 | 2.39 2 | 5.97 14 | 18.1 16 | 5.67 18 | 3.57 19 | 4.92 22 | 3.43 22 | 4.01 13 | 9.84 16 | 3.44 11 |
| Occlusion bounds [26] | 11.2 | 4.42 11 | 12.4 13 | 3.90 12 | 3.86 13 | 13.2 13 | 3.32 14 | 5.00 7 | 13.0 6 | 3.30 12 | 4.45 14 | 20.7 14 | 2.37 11 | 3.84 17 | 4.67 16 | 4.39 23 | 3.75 3 | 15.9 11 | 3.30 6 | 2.19 6 | 4.00 12 | 1.17 2 | 4.33 15 | 9.20 11 | 3.19 8 |
| Rannacher [27] | 11.7 | 4.13 9 | 11.0 9 | 3.61 11 | 3.39 11 | 12.3 11 | 2.80 11 | 7.26 17 | 17.4 19 | 3.59 16 | 4.40 13 | 23.1 20 | 2.24 9 | 3.43 6 | 4.54 9 | 2.56 5 | 5.41 12 | 18.5 17 | 4.23 10 | 2.92 15 | 3.91 9 | 2.82 15 | 3.45 7 | 9.14 10 | 3.27 9 |
| Brox et al. [5] | 11.9 | 4.44 13 | 12.4 13 | 4.22 18 | 3.72 12 | 13.5 14 | 3.06 12 | 4.97 6 | 13.3 8 | 3.11 10 | 4.58 16 | 22.0 18 | 2.37 11 | 3.79 14 | 4.60 12 | 4.33 22 | 3.91 5 | 17.0 15 | 3.45 7 | 2.22 7 | 3.79 7 | 1.19 3 | 4.62 16 | 10.0 17 | 3.38 10 |
| Multicue MRF [21] | 11.9 | 4.50 15 | 10.1 5 | 4.18 17 | 2.52 3 | 7.07 1 | 2.36 7 | 3.09 1 | 7.41 1 | 2.36 4 | 4.46 15 | 20.8 15 | 2.73 16 | 3.51 8 | 4.11 3 | 4.06 18 | 6.08 16 | 15.6 10 | 5.40 16 | 5.25 29 | 5.36 24 | 9.02 29 | 3.63 9 | 8.39 8 | 4.15 16 |
| F-TV-L1 [15] | 12.6 | 5.44 18 | 12.5 16 | 5.69 21 | 5.46 18 | 15.0 18 | 4.03 17 | 7.48 18 | 16.3 14 | 3.42 14 | 5.08 18 | 23.3 21 | 2.81 17 | 3.42 5 | 4.34 4 | 3.03 8 | 4.05 6 | 15.1 9 | 3.18 5 | 2.43 9 | 3.92 10 | 1.87 7 | 3.90 11 | 9.35 14 | 2.61 4 |
| CBF [12] | 13.0 | 3.88 6 | 10.2 7 | 3.50 10 | 4.60 17 | 11.3 9 | 5.06 18 | 5.43 9 | 13.1 7 | 3.39 13 | 4.09 10 | 21.2 16 | 2.16 7 | 3.80 15 | 4.72 19 | 3.52 12 | 4.33 7 | 14.4 7 | 3.01 4 | 4.97 27 | 5.51 26 | 4.93 27 | 3.99 12 | 9.27 13 | 3.91 15 |
| Dynamic MRF [7] | 15.2 | 4.58 16 | 12.4 13 | 4.14 16 | 3.25 10 | 13.9 15 | 2.27 5 | 6.02 10 | 16.8 16 | 2.36 4 | 4.39 12 | 22.6 19 | 2.51 14 | 3.61 10 | 4.55 10 | 3.46 10 | 6.81 22 | 22.2 24 | 6.78 22 | 2.41 8 | 3.48 4 | 3.69 23 | 9.26 27 | 17.8 28 | 10.2 27 |
| Fusion [6] | 15.2 | 4.43 12 | 13.7 19 | 4.08 15 | 2.47 1 | 8.91 2 | 2.24 4 | 3.70 3 | 9.68 3 | 3.12 11 | 3.68 7 | 19.8 11 | 2.54 15 | 4.26 23 | 5.16 22 | 4.31 21 | 6.32 18 | 16.8 14 | 6.15 20 | 4.55 25 | 5.78 27 | 3.10 20 | 7.12 24 | 13.6 24 | 7.86 25 |
| Second-order prior [8] | 15.5 | 4.03 8 | 11.6 11 | 3.35 7 | 3.88 14 | 14.0 16 | 3.08 13 | 7.21 16 | 17.6 20 | 3.57 15 | 4.14 11 | 19.9 12 | 2.31 10 | 3.66 11 | 4.86 21 | 2.73 6 | 7.32 23 | 21.2 21 | 6.76 21 | 4.02 23 | 4.58 18 | 4.01 25 | 4.27 14 | 10.4 19 | 5.12 18 |
| SegOF [10] | 15.9 | 5.85 19 | 13.5 18 | 3.98 13 | 7.40 20 | 14.9 17 | 8.13 25 | 8.55 20 | 17.3 18 | 9.01 20 | 6.50 21 | 18.1 7 | 5.14 21 | 3.90 20 | 4.53 8 | 4.81 26 | 6.57 21 | 21.7 23 | 6.81 23 | 1.65 1 | 3.49 5 | 1.08 1 | 3.71 10 | 9.23 12 | 3.63 12 |
| Learning Flow [11] | 18.1 | 4.23 10 | 11.7 12 | 3.41 8 | 4.16 15 | 15.3 19 | 3.42 15 | 6.78 13 | 16.9 17 | 3.83 18 | 6.41 20 | 25.3 23 | 4.25 19 | 4.66 25 | 6.01 28 | 4.00 17 | 6.33 20 | 20.7 20 | 5.30 13 | 3.09 17 | 4.84 20 | 2.91 18 | 7.08 23 | 15.0 25 | 5.27 20 |
| GraphCuts [14] | 19.0 | 6.25 20 | 14.3 20 | 5.53 20 | 8.60 22 | 20.1 25 | 6.61 20 | 7.91 19 | 15.4 11 | 10.9 21 | 4.88 17 | 19.0 9 | 3.05 18 | 3.78 13 | 4.71 17 | 3.94 16 | 8.74 26 | 16.4 13 | 5.39 15 | 4.04 24 | 4.87 21 | 4.85 26 | 6.35 21 | 12.2 20 | 6.05 22 |
| Filter Flow [20] | 19.4 | 6.48 21 | 14.6 21 | 4.96 19 | 5.73 19 | 15.7 20 | 5.07 19 | 10.1 22 | 18.6 21 | 14.3 25 | 9.04 25 | 23.3 21 | 7.80 25 | 3.98 21 | 4.71 17 | 4.21 20 | 5.86 13 | 15.0 8 | 5.41 17 | 4.98 28 | 6.87 29 | 2.78 14 | 4.82 18 | 8.66 9 | 3.65 13 |
| SPSA-learn [13] | 20.2 | 6.84 23 | 16.7 24 | 6.74 22 | 8.47 21 | 19.4 23 | 7.49 22 | 12.5 23 | 23.1 24 | 13.1 24 | 8.40 24 | 25.8 24 | 7.08 24 | 3.87 19 | 4.66 15 | 4.10 19 | 6.32 18 | 18.8 18 | 6.89 24 | 2.56 10 | 3.85 8 | 1.79 5 | 7.29 25 | 12.5 22 | 7.47 24 |
| Black & Anandan [4] | 20.6 | 6.81 22 | 15.4 22 | 7.43 23 | 8.77 23 | 19.5 24 | 7.35 21 | 13.0 25 | 22.9 23 | 12.5 22 | 8.29 23 | 26.1 25 | 6.77 23 | 4.18 22 | 5.28 23 | 3.69 14 | 6.19 17 | 20.0 19 | 5.34 14 | 3.63 21 | 5.05 23 | 1.79 5 | 6.45 22 | 12.2 20 | 5.17 19 |
| 2D-CLG [1] | 20.8 | 10.1 26 | 22.6 28 | 7.59 24 | 9.84 26 | 16.9 21 | 11.1 27 | 16.9 27 | 28.2 27 | 18.8 29 | 14.1 27 | 31.1 27 | 13.1 27 | 3.86 18 | 4.62 14 | 4.53 24 | 5.98 15 | 21.2 21 | 5.97 19 | 1.76 2 | 3.14 1 | 1.46 4 | 6.29 20 | 12.9 23 | 5.81 21 |
| GroupFlow [9] | 21.1 | 8.00 24 | 18.6 25 | 8.09 25 | 11.1 27 | 23.7 28 | 10.3 26 | 12.6 24 | 25.6 25 | 12.8 23 | 5.84 19 | 20.3 13 | 4.39 20 | 4.69 26 | 5.81 25 | 3.67 13 | 9.29 27 | 22.4 25 | 10.1 27 | 2.11 3 | 3.99 11 | 2.29 12 | 5.75 19 | 10.0 17 | 7.39 23 |
| Bipartite [30] | 22.3 | 12.8 29 | 16.4 23 | 9.33 27 | 8.97 24 | 19.1 22 | 7.87 24 | 9.52 21 | 19.6 22 | 7.50 19 | 7.17 22 | 16.4 5 | 5.81 22 | 5.96 29 | 6.24 29 | 7.02 28 | 15.2 30 | 23.0 26 | 16.8 30 | 13.3 31 | 13.4 31 | 8.86 28 | 2.85 4 | 8.16 6 | 2.37 3 |
| Horn & Schunck [3] | 24.8 | 8.01 25 | 19.9 26 | 8.38 26 | 9.13 25 | 23.2 27 | 7.71 23 | 14.2 26 | 25.9 26 | 14.6 26 | 12.4 26 | 30.6 26 | 11.3 26 | 4.64 24 | 5.64 24 | 4.60 25 | 8.21 25 | 24.4 28 | 8.45 25 | 4.01 22 | 5.41 25 | 1.95 10 | 9.16 26 | 17.5 26 | 8.86 26 |
| TI-DOFE [28] | 26.2 | 13.4 30 | 23.2 29 | 16.5 30 | 16.5 29 | 24.1 29 | 18.2 29 | 20.2 30 | 31.1 30 | 20.6 30 | 19.9 29 | 32.9 29 | 20.8 29 | 4.89 27 | 5.90 26 | 5.54 27 | 8.04 24 | 23.9 27 | 8.81 26 | 2.97 16 | 4.34 13 | 1.88 8 | 10.9 28 | 17.7 27 | 11.9 28 |
| STOB [22] | 27.4 | 11.6 28 | 26.0 31 | 14.6 29 | 15.3 28 | 25.0 30 | 17.5 28 | 17.8 29 | 30.1 29 | 18.1 28 | 25.4 31 | 33.6 30 | 28.0 31 | 5.25 28 | 5.90 26 | 7.03 29 | 10.3 28 | 27.4 30 | 10.6 28 | 2.89 14 | 4.47 16 | 2.94 19 | 14.9 29 | 20.7 29 | 18.8 29 |
| FOLKI [16] | 29.1 | 10.5 27 | 25.6 30 | 11.9 28 | 20.9 30 | 26.2 31 | 26.1 30 | 17.6 28 | 31.1 30 | 16.5 27 | 15.4 28 | 32.6 28 | 16.0 28 | 6.16 30 | 6.53 30 | 9.07 30 | 12.2 29 | 29.7 31 | 13.0 29 | 4.67 26 | 5.83 28 | 9.41 30 | 18.2 30 | 22.8 30 | 25.1 30 |
| Pyramid LK [2] | 30.2 | 13.9 31 | 20.9 27 | 21.4 31 | 24.1 31 | 23.1 26 | 30.2 31 | 20.9 31 | 29.5 28 | 21.9 31 | 22.2 30 | 34.6 31 | 25.0 30 | 18.7 31 | 23.1 31 | 20.2 31 | 21.2 31 | 24.5 29 | 21.0 31 | 6.41 30 | 7.02 30 | 10.8 31 | 25.6 31 | 31.5 31 | 34.5 31 |
|
Color encoding of flow vectors ![]() |
Army - Ground-truth flow
|
![]() |
![]() |
| Method | time* | frames | color | Reference and notes | |
| [1] 2D-CLG | 844 | 2 | gray | The 2D-CLG method by Bruhn et al. as implemented by Stefan Roth. [A. Bruhn, J. Weickert, and C. Schnörr. Lucas/Kanade meets Horn/Schunck: combining local and global optic flow methods. IJCV 63(3), 2005.] Parameters were set to match the published performance on Yosemite sequence, which may not be optimal for other sequences. | |
| [2] Pyramid LK | 12 | 2 | color | A modification of Bouguet's pyramidal implementation of Lucas-Kanade. | |
| [3] Horn & Schunck | 49 | 2 | gray | A modern Matlab implementation of the Horn & Schunck method by Deqing Sun. Parameters set to optimize AAE on all training data. | |
| [4] Black & Anandan | 328 | 2 | gray | A modern Matlab implementation of the Black & Anandan method by Deqing Sun. | |
| [5] Brox et al. | 18 | 2 | color | T. Brox, A. Bruhn, N. Papenberg, and J. Weickert. High accuracy optical flow estimation based on a theory for warping. ECCV 2004. (Improved using separate robust functions as proposed in A. Bruhn and J. Weickert, Towards ultimate motion estimation, ICCV 2005; improved by training on the training set.) | |
| [6] Fusion | 2,666 | 2 | color | V. Lempitsky, S. Roth, and C. Rother. Discrete-continuous optimization for optical flow estimation. CVPR 2008. | |
| [7] Dynamic MRF | 366 | 2 | gray | B. Glocker, N. Paragios, N. Komodakis, G. Tziritas, and N. Navab. Optical flow estimation with uncertainties through dynamic MRFs. CVPR 2008. (Method improved since publication.) | |
| [8] Second-order prior | 14 | 2 | gray | W. Trobin, T. Pock, D. Cremers, and H. Bischof. An unbiased second-order prior for high-accuracy motion estimation. DAGM 2008. (Method improved since publication; for details see W. Trobin, Ph.D. thesis, 2009.) | |
| [9] GroupFlow | 600 | 2 | gray | X. Ren. Local Grouping for Optical Flow. CVPR 2008. | |
| [10] SegOF | 60 | 2 | color | L. Xu, J. Chen, and J. Jia. Segmentation based variational model for accurate optical flow estimation. ECCV 2008. | |
| [11] Learning Flow | 825 | 2 | gray | D. Sun, S. Roth, J.P. Lewis, and M. Black. Learning optical flow (SRF-LFC). ECCV 2008. | |
| [12] CBF | 69 | 2 | color | W. Trobin, T. Pock, D. Cremers, and H. Bischof. Continuous energy minimization via repeated binary fusion. ECCV 2008. (Method improved since publication; for details see W. Trobin, Ph.D. thesis, 2009.) | |
| [13] SPSA-learn | 200 | 2 | color | Y. Li and D. Huttenlocher. Learning for optical flow using stochastic optimization. ECCV 2008. | |
| [14] GraphCuts | 1,200 | 2 | color | T. Cooke. Two applications of graph-cuts to image processing. DICTA 2008. | |
| [15] F-TV-L1 | 8 | 2 | gray | A. Wedel, T. Pock, J. Braun, U. Franke, and D. Cremers. Duality TV-L1 flow with fundamental matrix prior. IVCNZ 2008. | |
| [16] FOLKI | 1.4 | 2 | gray | G. Le Besnerais and F. Champagnat. Dense optical flow by iterative local window registration. ICIP 2005. | |
| [17] TV-L1-improved | 2.9 | 2 | gray | A. Wedel, T. Pock, C. Zach, H. Bischof, and D. Cremers. An improved algorithm for TV-L1 optical flow computation. Proceedings of the Dagstuhl Visual Motion Analysis Workshop 2008. Code at GPU4Vision. | |
| [18] DPOF | 261 | 2 | color | C. Lei and Y.-H. Yang. Optical flow estimation on coarse-to-fine region-trees using discrete optimization. ICCV 2009. | |
| [19] Spatially variant | 2,100 | 2 | color | Anonymous. Optical flow estimation with spatially-variant smoothness constraint. ICCV 2009 submission 1860. | |
| [20] Filter Flow | 34,000 | 2 | color | S. Seitz and S. Baker. Filter flow. ICCV 2009. | |
| [21] Multicue MRF | 13,240 | 2 | color | Anonymous. Optical flow estimation using discrete optimization with multi-cue weighted correlation and occlusion handling. ICCV 2009 submission 766. | |
| [22] STOB | 1,080 | 2 | gray | Anonymous. Stochastic uncertainty models for motion estimation. ICCV 2009 submission 1000. | |
| [23] Adaptive | 9.2 | 2 | gray | A. Wedel, D. Cremers, T. Pock, and H. Bischof. Structure- and motion-adaptive regularization for high accuracy optic flow. ICCV 2009. | |
| [24] Complementary OF | 44 | 2 | color | H. Zimmer, A. Bruhn, J. Weickert, L. Valgaerts, A. Salgado, B. Rosenhahn, and H.-P. Seidel. Complementary optic flow. EMMCVPR 2009. | |
| [25] Aniso. Huber-L1 | 2 | 2 | gray | M. Werlberger, W. Trobin, T. Pock, A. Wedel, D. Cremers, and H. Bischof. Anisotropic Huber-L1 optical flow. BMVC 2009. Code at GPU4Vision. | |
| [26] Occlusion bounds | 300 | 3 | color | Anonymous. Occlusion boundaries. NIPS 2009 submission 245. | |
| [27] Rannacher | 0.12 | 2 | gray | J. Rannacher. Realtime 3D motion estimation on graphics hardware. Bachelor thesis, Heidelberg University, 2009. | |
| [28] TI-DOFE | 260 | 2 | gray | C. Cassisa, S. Simoens, and V. Prinet. Two-frame optical flow formulation in an unwarped multiresolution scheme. CIARP 2009. | |
| [29] NL-TV-NCC | 20 | 2 | color | Anonymous. Motion estimation with non-local total variation regularization. CVPR 2010 submission 778. | |
| [30] Bipartite | 120 | 2 | gray | Anonymous. Dynamic bipartite matching. CVPR 2010 submission 69. | |
| [31] Classic+Area | 791 | 2 | gray | Anonymous. Secrets of optical flow estimation and their principles. CVPR 2010 submission 477. |