Fully-connected CRFs with non-parametric pairwise potential

N.D.F. Campbell, K. Subr, J. Kautz

Research output: Contribution to journalArticle

18 Citations (Scopus)

Abstract

Conditional Random Fields (CRFs) are used for diverse tasks, ranging from image denoising to object recognition. For images, they are commonly defined as a graph with nodes corresponding to individual pixels and pairwise links that connect nodes to their immediate neighbors. Recent work has shown that fully-connected CRFs, where each node is connected to every other node, can be solved efficiently under the restriction that the pairwise term is a Gaussian kernel over a Euclidean feature space. In this paper, we generalize the pairwise terms to a non-linear dissimilarity measure that is not required to be a distance metric. To this end, we propose a density estimation technique to derive conditional pairwise potentials in a non-parametric manner. We then use an efficient embedding technique to estimate an approximate Euclidean feature space for these potentials, in which the pairwise term can still be expressed as a Gaussian kernel. We demonstrate that the use of non-parametric models for the pairwise interactions, conditioned on the input data, greatly increases expressive power whilst maintaining efficient inference.
Original languageEnglish
Pages (from-to)1658-1665
Number of pages8
JournalIEEE Computer Society Conference on Computer Vision and Pattern Recognition
DOIs
Publication statusPublished - 1 Jan 2013

Fingerprint

Image denoising
Object recognition
Pixels

Cite this

Fully-connected CRFs with non-parametric pairwise potential. / Campbell, N.D.F.; Subr, K.; Kautz, J.

In: IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 01.01.2013, p. 1658-1665.

Research output: Contribution to journalArticle

@article{0657e3cf76f04496ac6aebbea9d7a6d6,
title = "Fully-connected CRFs with non-parametric pairwise potential",
abstract = "Conditional Random Fields (CRFs) are used for diverse tasks, ranging from image denoising to object recognition. For images, they are commonly defined as a graph with nodes corresponding to individual pixels and pairwise links that connect nodes to their immediate neighbors. Recent work has shown that fully-connected CRFs, where each node is connected to every other node, can be solved efficiently under the restriction that the pairwise term is a Gaussian kernel over a Euclidean feature space. In this paper, we generalize the pairwise terms to a non-linear dissimilarity measure that is not required to be a distance metric. To this end, we propose a density estimation technique to derive conditional pairwise potentials in a non-parametric manner. We then use an efficient embedding technique to estimate an approximate Euclidean feature space for these potentials, in which the pairwise term can still be expressed as a Gaussian kernel. We demonstrate that the use of non-parametric models for the pairwise interactions, conditioned on the input data, greatly increases expressive power whilst maintaining efficient inference.",
author = "N.D.F. Campbell and K. Subr and J. Kautz",
year = "2013",
month = "1",
day = "1",
doi = "10.1109/CVPR.2013.217",
language = "English",
pages = "1658--1665",
journal = "IEEE Computer Society Conference on Computer Vision and Pattern Recognition",
issn = "1063-6919",
publisher = "IEEE",

}

TY - JOUR

T1 - Fully-connected CRFs with non-parametric pairwise potential

AU - Campbell, N.D.F.

AU - Subr, K.

AU - Kautz, J.

PY - 2013/1/1

Y1 - 2013/1/1

N2 - Conditional Random Fields (CRFs) are used for diverse tasks, ranging from image denoising to object recognition. For images, they are commonly defined as a graph with nodes corresponding to individual pixels and pairwise links that connect nodes to their immediate neighbors. Recent work has shown that fully-connected CRFs, where each node is connected to every other node, can be solved efficiently under the restriction that the pairwise term is a Gaussian kernel over a Euclidean feature space. In this paper, we generalize the pairwise terms to a non-linear dissimilarity measure that is not required to be a distance metric. To this end, we propose a density estimation technique to derive conditional pairwise potentials in a non-parametric manner. We then use an efficient embedding technique to estimate an approximate Euclidean feature space for these potentials, in which the pairwise term can still be expressed as a Gaussian kernel. We demonstrate that the use of non-parametric models for the pairwise interactions, conditioned on the input data, greatly increases expressive power whilst maintaining efficient inference.

AB - Conditional Random Fields (CRFs) are used for diverse tasks, ranging from image denoising to object recognition. For images, they are commonly defined as a graph with nodes corresponding to individual pixels and pairwise links that connect nodes to their immediate neighbors. Recent work has shown that fully-connected CRFs, where each node is connected to every other node, can be solved efficiently under the restriction that the pairwise term is a Gaussian kernel over a Euclidean feature space. In this paper, we generalize the pairwise terms to a non-linear dissimilarity measure that is not required to be a distance metric. To this end, we propose a density estimation technique to derive conditional pairwise potentials in a non-parametric manner. We then use an efficient embedding technique to estimate an approximate Euclidean feature space for these potentials, in which the pairwise term can still be expressed as a Gaussian kernel. We demonstrate that the use of non-parametric models for the pairwise interactions, conditioned on the input data, greatly increases expressive power whilst maintaining efficient inference.

UR - http://www.scopus.com/inward/record.url?scp=84887355384&partnerID=8YFLogxK

UR - http://dx.doi.org/10.1109/CVPR.2013.217

U2 - 10.1109/CVPR.2013.217

DO - 10.1109/CVPR.2013.217

M3 - Article

AN - SCOPUS:84887355384

SP - 1658

EP - 1665

JO - IEEE Computer Society Conference on Computer Vision and Pattern Recognition

JF - IEEE Computer Society Conference on Computer Vision and Pattern Recognition

SN - 1063-6919

ER -