This page displays the lattice detection results obtained using the method
(and the MATLAB code) developed by Hays et al.:
Discovering Texture Regularity as a Higher-Order Correspondence Problem
James Hays, Marius Leordeanu, Alexei Efros, and Yanxi Liu.
European Conference on Computer Vision (ECCV) 2006.
The table below shows example images that we analyzed using the original MATLAB
with unmodified parameters as provided by the authors. Please note, that we
deliberately picked input examples which we believe to be difficult for an automatic
lattice detection algorithm.
To produce the results below we copied input images to the 'test_set' directory and ran
'directory_batch.m' script. In the 'outputs' directory we examined *a_scores*.txt file
and picked
the best (minimal) a-score among all iterations and all seeds. The images
corresponding to the detected lattice with the best a-score are shown below.
Note that we modified 'texture_match.m' script to have the 'warp_image' flag set to 1,
and we print the image warped to lattice shown in the third column (OUTPUT: warp).
In the table below, the first column (INPUT) shows the input image,
the second column (OUTPUT: warp) shows the content of 'image_warped' variable,
the third column (OUTPUT: lattice), shows the *lattice_on_original_image*.bmp.
INPUT | OUTPUT: warp | OUTPUT: lattice |
---|---|---|
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
|
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |
![]() |