Method

Robust Local Optical Flow Estimation using Bilinear Equations for Sparse Motion Estimation [BERLOF]
http://www.nue.tu-berlin.de/menue/forschung/projekte/rlof/

Submitted on 16 May. 2013 13:49 by
Tobias Senst (Technische Universität Berlin)

Running time:0.231 s
Environment:GPU @ 700 Mhz (C/C++) GeForce GTX 680

Method Description:
The Method is based on the Robust Local Optical Flow method. A new iterative refinement scheme based on a system of bilinear equations is used that lead to direct solutions.
Parameters:
Level = 4
max. no. Iterations = 30
Window-Size = 15x15
\sigma = (16, 50)
Confidence-Type = Forward/Backward Motion
Confidence-Threshold = 0.3

Latex Bibtex:
@INPROCEEDINGS{Senst_ICIP_2013,
AUTHOR = {Tobias Senst and Jonas Geistert and Ivo Keller and Thomas Sikora},
TITLE = {Robust Local Optical Flow Estimation using Bilinear Equations for Sparse Motion Estimation},
BOOKTITLE = {20th IEEE International Conference on Image Processing},
YEAR = {2013},
MONTH = sep,
ADDRESS = {Melbourne, Australia},
ABSTRACT = {This article presents a theoretical framework to decrease the computation effort of the Robust Local Optical Flow method which is based on the Lucas Kanade method. We show mathematically,how to transform the iterative scheme of the feature tracker into a system of bilinear equations and thus estimate the motion vectors directly by analyzing its zeros. Furthermore, we show that it is possible to parallelise our approach efficiently on a GPU, thus, outperforming the current OpenCV-OpenCL implementation of the pyramidal Lucas Kanade method in terms of runtime and accuracy. Finally, an evaluation is given for the Middlebury Optical Flow and the KITTI datasets.}
}

Detailed Results

This page provides detailed results for the method(s) selected. For each of the first 20 test images, the number of erroneous pixels at all thresholds is depicted in the table. Underneath, the left input image, the disparity / end-point error map and the estimated (and interpolated) disparity / optical flow map are shown. The error map scales linearly between 0 (black) and >=5 (white) pixels error. Red denotes all occluded pixels, falling outside the image boundaries. The false color map is scaled to the largest ground truth disparity / flow value.

Test Set Average

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 43.85 % 50.69 % 8.5 px 16.2 px
3 pixels 37.66 % 45.27 % 8.5 px 16.2 px
4 pixels 33.67 % 41.73 % 8.5 px 16.2 px
5 pixels 30.69 % 39.07 % 8.5 px 16.2 px
This table as LaTeX

Reflective Regions

This table as LaTeX

Test Image 0

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 36.07 % 45.51 % 5.8 px 12.1 px
3 pixels 31.94 % 41.96 % 5.8 px 12.1 px
4 pixels 28.99 % 39.40 % 5.8 px 12.1 px
5 pixels 26.70 % 37.41 % 5.8 px 12.1 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 1

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 34.19 % 42.95 % 3.5 px 7.5 px
3 pixels 27.11 % 36.58 % 3.5 px 7.5 px
4 pixels 22.65 % 32.51 % 3.5 px 7.5 px
5 pixels 19.46 % 29.49 % 3.5 px 7.5 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 2

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 46.12 % 54.71 % 7.9 px 15.5 px
3 pixels 41.32 % 50.64 % 7.9 px 15.5 px
4 pixels 38.19 % 47.99 % 7.9 px 15.5 px
5 pixels 35.83 % 45.96 % 7.9 px 15.5 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 3

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 55.58 % 62.90 % 9.2 px 16.1 px
3 pixels 47.84 % 56.41 % 9.2 px 16.1 px
4 pixels 42.37 % 51.77 % 9.2 px 16.1 px
5 pixels 38.11 % 48.13 % 9.2 px 16.1 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 4

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 37.08 % 41.64 % 4.5 px 6.5 px
3 pixels 29.37 % 34.15 % 4.5 px 6.5 px
4 pixels 24.58 % 29.50 % 4.5 px 6.5 px
5 pixels 21.19 % 26.17 % 4.5 px 6.5 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 5

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 45.54 % 53.74 % 9.5 px 21.0 px
3 pixels 38.04 % 47.17 % 9.5 px 21.0 px
4 pixels 33.88 % 43.46 % 9.5 px 21.0 px
5 pixels 31.18 % 41.01 % 9.5 px 21.0 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 6

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 39.48 % 47.38 % 6.4 px 11.6 px
3 pixels 33.44 % 41.99 % 6.4 px 11.6 px
4 pixels 29.45 % 38.34 % 6.4 px 11.6 px
5 pixels 26.48 % 35.49 % 6.4 px 11.6 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 7

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 36.62 % 47.12 % 6.8 px 17.0 px
3 pixels 30.38 % 41.82 % 6.8 px 17.0 px
4 pixels 26.32 % 38.35 % 6.8 px 17.0 px
5 pixels 23.48 % 35.81 % 6.8 px 17.0 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 8

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 41.27 % 51.78 % 9.1 px 22.4 px
3 pixels 34.02 % 45.79 % 9.1 px 22.4 px
4 pixels 29.59 % 42.10 % 9.1 px 22.4 px
5 pixels 26.27 % 39.34 % 9.1 px 22.4 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 9

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 63.72 % 69.33 % 11.0 px 19.9 px
3 pixels 56.01 % 62.78 % 11.0 px 19.9 px
4 pixels 50.33 % 57.92 % 11.0 px 19.9 px
5 pixels 45.83 % 54.08 % 11.0 px 19.9 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 10

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 50.80 % 59.46 % 9.0 px 22.2 px
3 pixels 43.88 % 53.74 % 9.0 px 22.2 px
4 pixels 38.96 % 49.67 % 9.0 px 22.2 px
5 pixels 35.11 % 46.49 % 9.0 px 22.2 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 11

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 47.50 % 55.60 % 5.3 px 12.0 px
3 pixels 37.31 % 46.90 % 5.3 px 12.0 px
4 pixels 30.99 % 41.48 % 5.3 px 12.0 px
5 pixels 26.64 % 37.63 % 5.3 px 12.0 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 12

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 50.09 % 57.98 % 10.3 px 21.6 px
3 pixels 43.29 % 52.21 % 10.3 px 21.6 px
4 pixels 38.58 % 48.18 % 10.3 px 21.6 px
5 pixels 34.85 % 45.01 % 10.3 px 21.6 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 13

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 45.37 % 51.98 % 10.2 px 16.4 px
3 pixels 40.98 % 48.10 % 10.2 px 16.4 px
4 pixels 38.23 % 45.69 % 10.2 px 16.4 px
5 pixels 35.99 % 43.71 % 10.2 px 16.4 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 14

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 67.93 % 73.38 % 19.7 px 30.3 px
3 pixels 63.27 % 69.51 % 19.7 px 30.3 px
4 pixels 59.94 % 66.73 % 19.7 px 30.3 px
5 pixels 57.18 % 64.42 % 19.7 px 30.3 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 15

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 60.54 % 67.28 % 9.1 px 20.8 px
3 pixels 52.97 % 60.86 % 9.1 px 20.8 px
4 pixels 47.61 % 56.24 % 9.1 px 20.8 px
5 pixels 43.60 % 52.70 % 9.1 px 20.8 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 16

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 14.48 % 22.89 % 3.9 px 8.2 px
3 pixels 12.97 % 21.44 % 3.9 px 8.2 px
4 pixels 12.05 % 20.54 % 3.9 px 8.2 px
5 pixels 11.30 % 19.76 % 3.9 px 8.2 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 17

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 41.14 % 50.26 % 4.7 px 14.5 px
3 pixels 32.66 % 43.03 % 4.7 px 14.5 px
4 pixels 27.71 % 38.78 % 4.7 px 14.5 px
5 pixels 24.27 % 35.80 % 4.7 px 14.5 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 18

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 25.39 % 33.00 % 5.3 px 9.4 px
3 pixels 20.55 % 28.61 % 5.3 px 9.4 px
4 pixels 18.42 % 26.63 % 5.3 px 9.4 px
5 pixels 17.32 % 25.54 % 5.3 px 9.4 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot



Test Image 19

Error Out-Noc Out-All Avg-Noc Avg-All
2 pixels 38.08 % 45.10 % 6.5 px 10.9 px
3 pixels 32.27 % 39.68 % 6.5 px 10.9 px
4 pixels 28.51 % 36.03 % 6.5 px 10.9 px
5 pixels 25.56 % 33.09 % 6.5 px 10.9 px
This table as LaTeX





This figure as: png eps pdf txt gnuplot





eXTReMe Tracker