Multiscale exemplar based texture synthesis by locally gaussian models

Lara RAAD; Agnès DESOLNEUX; Jean Michel MOREL

doi:10.1007/978-3-319-25751-8_52

Multiscale exemplar based texture synthesis by locally gaussian models

Lara RAAD^*
, Agnès DESOLNEUX
, Jean Michel MOREL

^*Corresponding author for this work

Research output: Book Chapters | Papers in Conference Proceedings › Conference paper (refereed) › Research › peer-review

4 Citations (Scopus)

Abstract

In exemplar based texture synthesis methods one of the major difficulties is to synthesize correctly the wide diversity of texture images. So far the proposed methods tend to have satisfying results for specific texture classes and fail for others. Statistics-based algorithms present good results when synthesizing textures that have few geometric structures and are able to preserve a complex statistical model of the sample texture. On the other hand, non-parametric patch-based methods have the ability to reproduce faithfully highly structured textures but lack a mechanism to preserve its global statistics. Furthermore, they are strongly dependent on a patch size that is decided manually. In this paper we propose a multiscale approach able to combine advantages of both strategies and avoid some of their drawbacks. The texture is modeled at each scale as a spatially variable Gaussian vector in the patch space, which allows to fix a patch size fairly independent of the texture.

Original language	English
Title of host publication	Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications, 20th Iberoamerican Congress, CIARP 2015, Proceedings
Editors	Alvaro PARDO, Josef KITTLER
Publisher	Springer, Cham
Pages	435-443
Number of pages	9
ISBN (Print)	9783319257501
DOIs	https://doi.org/10.1007/978-3-319-25751-8_52
Publication status	Published - 2015
Externally published	Yes
Event	20th Iberoamerican Congress on on Pattern Recognition, CIARP 2015 - Montevideo, Uruguay Duration: 9 Nov 2015 → 12 Nov 2015

Publication series

Name	Lecture Notes in Computer Science
Volume	9423
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	20th Iberoamerican Congress on on Pattern Recognition, CIARP 2015
Country/Territory	Uruguay
City	Montevideo
Period	9/11/15 → 12/11/15

Bibliographical note

Publisher Copyright:
© Springer International Publishing Switzerland 2015.

Funding

Work partly founded by the European Research Council (advanced grant Twelve Labours) and the Office of Naval research (ONR grant N00014-14-1-0023).

Keywords

Locally gaussian
Multiscale
Patch size
Texture synthesis

Access to Document

10.1007/978-3-319-25751-8_52

Cite this

RAAD, L., DESOLNEUX, A., & MOREL, J. M. (2015). Multiscale exemplar based texture synthesis by locally gaussian models. In A. PARDO, & J. KITTLER (Eds.), Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications, 20th Iberoamerican Congress, CIARP 2015, Proceedings (pp. 435-443). (Lecture Notes in Computer Science; Vol. 9423). Springer, Cham. https://doi.org/10.1007/978-3-319-25751-8_52

RAAD, Lara ; DESOLNEUX, Agnès ; MOREL, Jean Michel. / Multiscale exemplar based texture synthesis by locally gaussian models. Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications, 20th Iberoamerican Congress, CIARP 2015, Proceedings . editor / Alvaro PARDO ; Josef KITTLER. Springer, Cham, 2015. pp. 435-443 (Lecture Notes in Computer Science).

@inproceedings{782f6baade5c4072a62b47b42235896e,

title = "Multiscale exemplar based texture synthesis by locally gaussian models",

abstract = "In exemplar based texture synthesis methods one of the major difficulties is to synthesize correctly the wide diversity of texture images. So far the proposed methods tend to have satisfying results for specific texture classes and fail for others. Statistics-based algorithms present good results when synthesizing textures that have few geometric structures and are able to preserve a complex statistical model of the sample texture. On the other hand, non-parametric patch-based methods have the ability to reproduce faithfully highly structured textures but lack a mechanism to preserve its global statistics. Furthermore, they are strongly dependent on a patch size that is decided manually. In this paper we propose a multiscale approach able to combine advantages of both strategies and avoid some of their drawbacks. The texture is modeled at each scale as a spatially variable Gaussian vector in the patch space, which allows to fix a patch size fairly independent of the texture.",

keywords = "Locally gaussian, Multiscale, Patch size, Texture synthesis",

author = "Lara RAAD and Agn{\`e}s DESOLNEUX and MOREL, \{Jean Michel\}",

note = "Publisher Copyright: {\textcopyright} Springer International Publishing Switzerland 2015.; 20th Iberoamerican Congress on on Pattern Recognition, CIARP 2015 ; Conference date: 09-11-2015 Through 12-11-2015",

year = "2015",

doi = "10.1007/978-3-319-25751-8\_52",

language = "English",

isbn = "9783319257501",

series = "Lecture Notes in Computer Science",

publisher = "Springer, Cham",

pages = "435--443",

editor = "Alvaro PARDO and Josef KITTLER",

booktitle = "Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications, 20th Iberoamerican Congress, CIARP 2015, Proceedings",

}

RAAD, L, DESOLNEUX, A & MOREL, JM 2015, Multiscale exemplar based texture synthesis by locally gaussian models. in A PARDO & J KITTLER (eds), Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications, 20th Iberoamerican Congress, CIARP 2015, Proceedings . Lecture Notes in Computer Science, vol. 9423, Springer, Cham, pp. 435-443, 20th Iberoamerican Congress on on Pattern Recognition, CIARP 2015, Montevideo, Uruguay, 9/11/15. https://doi.org/10.1007/978-3-319-25751-8_52

Multiscale exemplar based texture synthesis by locally gaussian models. / RAAD, Lara; DESOLNEUX, Agnès; MOREL, Jean Michel.
Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications, 20th Iberoamerican Congress, CIARP 2015, Proceedings . ed. / Alvaro PARDO; Josef KITTLER. Springer, Cham, 2015. p. 435-443 (Lecture Notes in Computer Science; Vol. 9423).

Research output: Book Chapters | Papers in Conference Proceedings › Conference paper (refereed) › Research › peer-review

TY - GEN

T1 - Multiscale exemplar based texture synthesis by locally gaussian models

AU - RAAD, Lara

AU - DESOLNEUX, Agnès

AU - MOREL, Jean Michel

N1 - Publisher Copyright: © Springer International Publishing Switzerland 2015.

PY - 2015

Y1 - 2015

N2 - In exemplar based texture synthesis methods one of the major difficulties is to synthesize correctly the wide diversity of texture images. So far the proposed methods tend to have satisfying results for specific texture classes and fail for others. Statistics-based algorithms present good results when synthesizing textures that have few geometric structures and are able to preserve a complex statistical model of the sample texture. On the other hand, non-parametric patch-based methods have the ability to reproduce faithfully highly structured textures but lack a mechanism to preserve its global statistics. Furthermore, they are strongly dependent on a patch size that is decided manually. In this paper we propose a multiscale approach able to combine advantages of both strategies and avoid some of their drawbacks. The texture is modeled at each scale as a spatially variable Gaussian vector in the patch space, which allows to fix a patch size fairly independent of the texture.

AB - In exemplar based texture synthesis methods one of the major difficulties is to synthesize correctly the wide diversity of texture images. So far the proposed methods tend to have satisfying results for specific texture classes and fail for others. Statistics-based algorithms present good results when synthesizing textures that have few geometric structures and are able to preserve a complex statistical model of the sample texture. On the other hand, non-parametric patch-based methods have the ability to reproduce faithfully highly structured textures but lack a mechanism to preserve its global statistics. Furthermore, they are strongly dependent on a patch size that is decided manually. In this paper we propose a multiscale approach able to combine advantages of both strategies and avoid some of their drawbacks. The texture is modeled at each scale as a spatially variable Gaussian vector in the patch space, which allows to fix a patch size fairly independent of the texture.

KW - Locally gaussian

KW - Multiscale

KW - Patch size

KW - Texture synthesis

UR - https://www.scopus.com/pages/publications/84983657038

U2 - 10.1007/978-3-319-25751-8_52

DO - 10.1007/978-3-319-25751-8_52

M3 - Conference paper (refereed)

AN - SCOPUS:84983657038

SN - 9783319257501

T3 - Lecture Notes in Computer Science

SP - 435

EP - 443

BT - Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications, 20th Iberoamerican Congress, CIARP 2015, Proceedings

A2 - PARDO, Alvaro

A2 - KITTLER, Josef

PB - Springer, Cham

T2 - 20th Iberoamerican Congress on on Pattern Recognition, CIARP 2015

Y2 - 9 November 2015 through 12 November 2015

ER -

RAAD L, DESOLNEUX A, MOREL JM. Multiscale exemplar based texture synthesis by locally gaussian models. In PARDO A, KITTLER J, editors, Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications, 20th Iberoamerican Congress, CIARP 2015, Proceedings . Springer, Cham. 2015. p. 435-443. (Lecture Notes in Computer Science). Epub 2015 Oct 25. doi: 10.1007/978-3-319-25751-8_52

Multiscale exemplar based texture synthesis by locally gaussian models

Abstract

Publication series

Conference

Bibliographical note

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this