In Defense of 3D-Label Stereo
(2013) IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2013 p.1730-1737- Abstract
- It is commonly believed that higher order smoothness
should be modeled using higher order interactions. For example,
2nd order derivatives for deformable (active) contours
are represented by triple cliques. Similarly, the 2nd
order regularization methods in stereo predominantly use
MRF models with scalar (1D) disparity labels and triple
clique interactions. In this paper we advocate a largely
overlooked alternative approach to stereo where 2nd order
surface smoothness is represented by pairwise interactions
with 3D-labels, e.g. tangent planes. This general paradigm
has been criticized due to perceived computational complexity
of optimization... (More) - It is commonly believed that higher order smoothness
should be modeled using higher order interactions. For example,
2nd order derivatives for deformable (active) contours
are represented by triple cliques. Similarly, the 2nd
order regularization methods in stereo predominantly use
MRF models with scalar (1D) disparity labels and triple
clique interactions. In this paper we advocate a largely
overlooked alternative approach to stereo where 2nd order
surface smoothness is represented by pairwise interactions
with 3D-labels, e.g. tangent planes. This general paradigm
has been criticized due to perceived computational complexity
of optimization in higher-dimensional label space.
Contrary to popular beliefs, we demonstrate that representing
2nd order surface smoothness with 3D labels leads
to simpler optimization problems with (nearly) submodular
pairwise interactions. Our theoretical and experimental results
demonstrate advantages over state-of-the-art methods
for 2nd order smoothness stereo. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/record/3914799
- author
- Olsson, Carl LU ; Ulén, Johannes LU and Boykov, Yuri
- organization
- publishing date
- 2013
- type
- Chapter in Book/Report/Conference proceeding
- publication status
- published
- subject
- host publication
- Computer Vision and Pattern Recognition (CVPR), 2013 IEEE Conference on
- pages
- 8 pages
- publisher
- IEEE - Institute of Electrical and Electronics Engineers Inc.
- conference name
- IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2013
- conference location
- Portland, Oregon, USA., United States
- conference dates
- 2013-06-23 - 2013-06-28
- external identifiers
-
- wos:000331094301100
- scopus:84887358167
- ISSN
- 2163-6648
- 1063-6919
- DOI
- 10.1109/CVPR.2013.226
- language
- English
- LU publication?
- yes
- id
- 0bbd40f9-5373-4cdf-a789-03f470b97e46 (old id 3914799)
- alternative location
- http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=6619070
- date added to LUP
- 2016-04-01 10:02:41
- date last changed
- 2024-07-28 11:27:38
@inproceedings{0bbd40f9-5373-4cdf-a789-03f470b97e46, abstract = {{It is commonly believed that higher order smoothness<br/><br> should be modeled using higher order interactions. For example,<br/><br> 2nd order derivatives for deformable (active) contours<br/><br> are represented by triple cliques. Similarly, the 2nd<br/><br> order regularization methods in stereo predominantly use<br/><br> MRF models with scalar (1D) disparity labels and triple<br/><br> clique interactions. In this paper we advocate a largely<br/><br> overlooked alternative approach to stereo where 2nd order<br/><br> surface smoothness is represented by pairwise interactions<br/><br> with 3D-labels, e.g. tangent planes. This general paradigm<br/><br> has been criticized due to perceived computational complexity<br/><br> of optimization in higher-dimensional label space.<br/><br> Contrary to popular beliefs, we demonstrate that representing<br/><br> 2nd order surface smoothness with 3D labels leads<br/><br> to simpler optimization problems with (nearly) submodular<br/><br> pairwise interactions. Our theoretical and experimental results<br/><br> demonstrate advantages over state-of-the-art methods<br/><br> for 2nd order smoothness stereo.}}, author = {{Olsson, Carl and Ulén, Johannes and Boykov, Yuri}}, booktitle = {{Computer Vision and Pattern Recognition (CVPR), 2013 IEEE Conference on}}, issn = {{2163-6648}}, language = {{eng}}, pages = {{1730--1737}}, publisher = {{IEEE - Institute of Electrical and Electronics Engineers Inc.}}, title = {{In Defense of 3D-Label Stereo}}, url = {{http://dx.doi.org/10.1109/CVPR.2013.226}}, doi = {{10.1109/CVPR.2013.226}}, year = {{2013}}, }