BA method, simple synthetic images
For simple synthetic images:
I0 =
0 0.5000 0
0 1.0000 0
0 0.5000 0
0.1000 0 0
0.0500 0 0
0.0500 0.1000 0
I1 =
0.5000 0 0
1.0000 0 0
0.5000 0 0
0 0.1000 0
0 0.0500 0
0 0.0500 0.1000
Running code by D.Sun. Disabled texture decomposition, disabled multiscale processing.
uv = estimate_flow_interface(I0, I1, 'classic-c-brightness', [], {'display', 1, 'pyramid_levels', 1, 'gnc_pyramid_levels', 1});
After first iteration ( loop for ignc = 1:this.gnc_iters in dsun_ijcv_flow_code\@ba_optical_flow\compute_flow.m):
After 2nd iteration:
After 3rd iteration:
I think, it is right answer.
uv(:,:,1) =
-0.9995 -0.9995 -0.9996
-0.9994 -0.9994 -0.9995
-0.9992 -0.9992 -0.9992
0.9992 0.9992 0.9992
0.9994 0.9994 0.9994
0.9995 0.9994 0.9994
uv(:,:,2) =
1.0e-003 *
-0.2083 -0.2080 -0.2080
-0.2080 -0.1848 -0.1848
-0.1861 -0.1616 -0.1848
-0.1507 -0.0860 -0.1183
-0.0860 -0.0633 -0.1183
-0.0396 -0.0396 -0.0860
BA method. Another pair of images
I0 =
0 0.5000 0
0 1.0000 0
0 0.5000 0
I1 =
0.5000 0 0
1.0000 0 0
0.5000 0 0
1st iteration
uv(:,:,1) =
-1 -1 -1
-1 -1 -1
-1 -1 -1
uv(:,:,2) =
1.0e-014 *
0.8281 0.8281 0.8281
0.8281 0.8281 0.8281
0.8281 0.8281 0.8281
2nd iteration
uv(:,:,1) =
-1.0000 -1.0000 -1.0000
-1.0000 -1.0000 -1.0000
-1.0000 -1.0000 -1.0000
uv(:,:,2) =
1.0e-013 *
-0.5965 -0.5965 -0.5965
-0.5965 -0.5965 -0.5965
-0.5965 -0.5965 -0.5965
3rd iteration
uv(:,:,1) =
-1.0000 -1.0000 -1.0000
-1.0000 -1.0000 -1.0000
-1.0000 -1.0000 -1.0000
uv(:,:,2) =
1.0e-011 *
0.5374 0.5374 0.5374
0.5374 0.5374 0.5374
0.5374 0.5374 0.5374
Visually the same:
HS method. Simple image
BA - is not the simpliest method. I found there is implmentation of Horn-Schunk.
I found that the results are really poor, when my “image elements” ar just on the border. So I extended image by zeros.
I use the following command:
uv = estimate_flow_interface(I0, I1, ...
'hs-brightness', [], ...
{'display', 1, 'pyramid_levels', 1, 'gnc_pyramid_levels', 1, ...
'pyramid_spacing', sqrt(2)});
I also use slightly modified code. I disabled automatic pyramid height calculation, so I set it manually. Input
I0 =
0 0.5000 0
0 1.0000 0
0 0.5000 0
1.0000 0 0
0.5000 0 0
0.5000 1.0000 0
I1 =
0.5000 0 0
1.0000 0 0
0.5000 0 0
0 1.0000 0
0 0.5000 0
0 0.5000 1.0000
Gives:
While
I0 =
0 0 0 0 0
0 0 0.5000 0 0
0 0 1.0000 0 0
0 0 0.5000 0 0
0 1.0000 0 0 0
0 0.5000 0 0 0
0 0.5000 1.0000 0 0
0 0 0 0 0
I1 =
0 0 0 0 0
0 0.5000 0 0 0
0 1.0000 0 0 0
0 0.5000 0 0 0
0 0 1.0000 0 0
0 0 0.5000 0 0
0 0 0.5000 1.0000 0
0 0 0 0 0
gives
HS method. Multiscale.
I0 = [ 0 0 0 0 0 0 0 0 0 0 0 0
3 0 0 0 0 0 5 5 0 0 0 0
3 0 0 0 0 0 1 3 0 0 0 0
0 0 0 1 0 0 2 4 0 0 0 0
0 0 0 2 0 0 0 0 0 0 0 0
5 2 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
5 5 0 0 0 0 0 0 0 0 0 0
5 5 0 0 0 0 0 0 0 0 0 0 ] / 5.;
I1 = [ 0 0 0 0 0 0 0 0 0 0 0 0
3 0 0 0 0 0 0 0 0 0 0 0
3 0 0 0 0 0 0 0 0 0 0 0
0 0 0 1 0 0 0 0 0 5 5 0
0 0 0 2 0 0 0 0 0 1 3 0
5 2 0 0 0 0 0 0 0 2 4 0
0 0 0 0 0 0 0 0 0 0 0 0
5 5 0 0 0 0 0 0 0 0 0 0
5 5 0 0 0 0 0 0 0 0 0 0 ] / 5.;
Here the shift between frames is much larger than a pixel and the size of objects. So OF fail without resolution scaling.
| 1 pyramid level  | 2 pyramid levels  | 3 pyramid levels  |