Image recognition with occlusions

Tyng Luh Liu, Mike Donahue, Davi Geiger, Robert Hummel

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

We study the problem of how to detect “interesting objects” appeared in a given image, I. Our approach is to treat it as a function approximation problem based on an over-redundant basis. Since the basis (a library of image templates) is over-redundant, there are infinitely many ways to decompose I. To select the “best” decomposition we first propose a global optimization procedure that considers a concave cost function derived from a “weighted Lpnorm” with 0 <p ≤ 1. This concave cost function selects as few coefficients as possible producing a sparse representation of the image and handle occlusions. However, it contains multiple local minima. We identify all local minima so that a global optimization is possible by visiting all of them. Secondly, because the number of local minima grows exponentially with the number of templates, we investigate a greedy “LpMatching Pursuit” strategy.

Original languageEnglish (US)
Title of host publicationComputer Vision – ECCV 1996 - 4th European Conference on Computer Vision, Proceedings
PublisherSpringer Verlag
Pages556-565
Number of pages10
Volume1064
ISBN (Print)3540611223, 9783540611226
StatePublished - 1996
Event4th European Conference on Computer Vision, ECCV 1996 - Cambridge, United Kingdom
Duration: Apr 15 1996Apr 18 1996

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume1064
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other4th European Conference on Computer Vision, ECCV 1996
CountryUnited Kingdom
CityCambridge
Period4/15/964/18/96

Fingerprint

Image recognition
Image Recognition
Global optimization
Occlusion
Local Minima
Cost functions
Concave function
Global Optimization
Cost Function
Template
Decompose
Sparse Representation
Lp-norm
Pursuit
Approximation Problem
Function Approximation
Decomposition
Coefficient

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Liu, T. L., Donahue, M., Geiger, D., & Hummel, R. (1996). Image recognition with occlusions. In Computer Vision – ECCV 1996 - 4th European Conference on Computer Vision, Proceedings (Vol. 1064, pp. 556-565). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1064). Springer Verlag.

Image recognition with occlusions. / Liu, Tyng Luh; Donahue, Mike; Geiger, Davi; Hummel, Robert.

Computer Vision – ECCV 1996 - 4th European Conference on Computer Vision, Proceedings. Vol. 1064 Springer Verlag, 1996. p. 556-565 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1064).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Liu, TL, Donahue, M, Geiger, D & Hummel, R 1996, Image recognition with occlusions. in Computer Vision – ECCV 1996 - 4th European Conference on Computer Vision, Proceedings. vol. 1064, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 1064, Springer Verlag, pp. 556-565, 4th European Conference on Computer Vision, ECCV 1996, Cambridge, United Kingdom, 4/15/96.
Liu TL, Donahue M, Geiger D, Hummel R. Image recognition with occlusions. In Computer Vision – ECCV 1996 - 4th European Conference on Computer Vision, Proceedings. Vol. 1064. Springer Verlag. 1996. p. 556-565. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
Liu, Tyng Luh ; Donahue, Mike ; Geiger, Davi ; Hummel, Robert. / Image recognition with occlusions. Computer Vision – ECCV 1996 - 4th European Conference on Computer Vision, Proceedings. Vol. 1064 Springer Verlag, 1996. pp. 556-565 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{350aeea847704eb5a8a90469522c9e36,
title = "Image recognition with occlusions",
abstract = "We study the problem of how to detect “interesting objects” appeared in a given image, I. Our approach is to treat it as a function approximation problem based on an over-redundant basis. Since the basis (a library of image templates) is over-redundant, there are infinitely many ways to decompose I. To select the “best” decomposition we first propose a global optimization procedure that considers a concave cost function derived from a “weighted Lpnorm” with 0 pMatching Pursuit” strategy.",
author = "Liu, {Tyng Luh} and Mike Donahue and Davi Geiger and Robert Hummel",
year = "1996",
language = "English (US)",
isbn = "3540611223",
volume = "1064",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "556--565",
booktitle = "Computer Vision – ECCV 1996 - 4th European Conference on Computer Vision, Proceedings",

}

TY - GEN

T1 - Image recognition with occlusions

AU - Liu, Tyng Luh

AU - Donahue, Mike

AU - Geiger, Davi

AU - Hummel, Robert

PY - 1996

Y1 - 1996

N2 - We study the problem of how to detect “interesting objects” appeared in a given image, I. Our approach is to treat it as a function approximation problem based on an over-redundant basis. Since the basis (a library of image templates) is over-redundant, there are infinitely many ways to decompose I. To select the “best” decomposition we first propose a global optimization procedure that considers a concave cost function derived from a “weighted Lpnorm” with 0 pMatching Pursuit” strategy.

AB - We study the problem of how to detect “interesting objects” appeared in a given image, I. Our approach is to treat it as a function approximation problem based on an over-redundant basis. Since the basis (a library of image templates) is over-redundant, there are infinitely many ways to decompose I. To select the “best” decomposition we first propose a global optimization procedure that considers a concave cost function derived from a “weighted Lpnorm” with 0 pMatching Pursuit” strategy.

UR - http://www.scopus.com/inward/record.url?scp=84957560018&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84957560018&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:84957560018

SN - 3540611223

SN - 9783540611226

VL - 1064

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 556

EP - 565

BT - Computer Vision – ECCV 1996 - 4th European Conference on Computer Vision, Proceedings

PB - Springer Verlag

ER -