Local Refinement for Stereo Regularization
(2014) 22nd International Conference on Pattern Recognition (ICPR 2014) p.4056-4061- Abstract
- Stereo matching is an inherently difficult problem due to ambiguous and noisy texture. The non-convexity and non- differentiability makes local linear (or quadratic) approximations poor, thereby preventing the use of standard local descent methods. Therefore recent methods are predominantly based on discretization and/or random sampling of some class of approximating surfaces (e.g. planes). While these methods are very efficient in generating a rough surface estimate, via either fusion of proposals or label propagation, the end result is usually not as smooth as desired. In this paper we show that, if the objective function is decomposed correctly, local refinement of candidate solutions can be performed using an ADMM approach. This allows... (More)
- Stereo matching is an inherently difficult problem due to ambiguous and noisy texture. The non-convexity and non- differentiability makes local linear (or quadratic) approximations poor, thereby preventing the use of standard local descent methods. Therefore recent methods are predominantly based on discretization and/or random sampling of some class of approximating surfaces (e.g. planes). While these methods are very efficient in generating a rough surface estimate, via either fusion of proposals or label propagation, the end result is usually not as smooth as desired. In this paper we show that, if the objective function is decomposed correctly, local refinement of candidate solutions can be performed using an ADMM approach. This allows searching over more general function classes, thereby resulting in visually more appealing smooth surface estimations. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/record/4777616
- author
- Olsson, Carl LU ; Ulén, Johannes LU and Eriksson, Anders
- organization
- publishing date
- 2014
- type
- Chapter in Book/Report/Conference proceeding
- publication status
- published
- subject
- host publication
- Pattern Recognition (ICPR), 2014 22nd International Conference on
- pages
- 6 pages
- publisher
- IEE
- conference name
- 22nd International Conference on Pattern Recognition (ICPR 2014)
- conference location
- Stockholm, Sweden
- conference dates
- 2014-08-24 - 2014-08-28
- external identifiers
-
- wos:000359818004032
- scopus:84919935988
- ISSN
- 1051-4651
- DOI
- 10.1109/ICPR.2014.695
- language
- English
- LU publication?
- yes
- id
- 8daec9d8-859e-4a19-9bd4-2fe42f26d207 (old id 4777616)
- alternative location
- http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=6977408
- date added to LUP
- 2016-04-01 13:43:44
- date last changed
- 2022-01-27 20:42:52
@inproceedings{8daec9d8-859e-4a19-9bd4-2fe42f26d207, abstract = {{Stereo matching is an inherently difficult problem due to ambiguous and noisy texture. The non-convexity and non- differentiability makes local linear (or quadratic) approximations poor, thereby preventing the use of standard local descent methods. Therefore recent methods are predominantly based on discretization and/or random sampling of some class of approximating surfaces (e.g. planes). While these methods are very efficient in generating a rough surface estimate, via either fusion of proposals or label propagation, the end result is usually not as smooth as desired. In this paper we show that, if the objective function is decomposed correctly, local refinement of candidate solutions can be performed using an ADMM approach. This allows searching over more general function classes, thereby resulting in visually more appealing smooth surface estimations.}}, author = {{Olsson, Carl and Ulén, Johannes and Eriksson, Anders}}, booktitle = {{Pattern Recognition (ICPR), 2014 22nd International Conference on}}, issn = {{1051-4651}}, language = {{eng}}, pages = {{4056--4061}}, publisher = {{IEE}}, title = {{Local Refinement for Stereo Regularization}}, url = {{http://dx.doi.org/10.1109/ICPR.2014.695}}, doi = {{10.1109/ICPR.2014.695}}, year = {{2014}}, }