Data-driven geometry-recovering mesh denoising

Jun WANG, Jin HUANG, Fu Lee WANG, Mingqiang WEI, Haoran XIE, Jing QIN

Research output: Journal PublicationsJournal Article (refereed)

4 Citations (Scopus)

Abstract

Depth cameras and 3D scanners significantly simplify the procedure of geometric modeling. 3D surfaces have become more widespread, leading to a great demand for noise removal with the expectation of the minimal disturbance of mesh geometry. We propose a novel two-step data-driven mesh denoising approach. The first step removes noise by learning normal variations from noisy models to their ground-truth counterparts. Unlike existing denoising methods, we present the second step to recover the mesh geometry lost in the first step. The second step understands the commonly used filters by learning the mapping from filtered models to their ground-truth counterparts. In addition, (1) to handle noise with large variations, we model normal estimation as a low-rank matrix recovery problem in similar-patch collaboration before the first-step learning; (2) to recover the real geometry of a denoised mesh, we reversely filter the denoised mesh to obtain more geometry cues before the second-step learning. The detailed quantitative and qualitative results on various data demonstrate that, our two-step learning algorithm competes favorably with the state-of-the-art methods in terms of mesh geometry preservation and noise-robustness.
Original languageEnglish
Pages (from-to)133-142
Number of pages10
JournalCAD Computer Aided Design
Volume114
Early online date14 May 2019
DOIs
Publication statusPublished - Sep 2019
Externally publishedYes

Fingerprint

Geometry
Learning algorithms
Cameras
Recovery

Keywords

  • Geometry-recovering
  • Low-rank matrix recovery
  • Mesh denoising
  • Normal variation learning
  • Reverse filter

Cite this

WANG, Jun ; HUANG, Jin ; WANG, Fu Lee ; WEI, Mingqiang ; XIE, Haoran ; QIN, Jing. / Data-driven geometry-recovering mesh denoising. In: CAD Computer Aided Design. 2019 ; Vol. 114. pp. 133-142.
@article{3b42a48faaf24e2dbd68cc11409bde46,
title = "Data-driven geometry-recovering mesh denoising",
abstract = "Depth cameras and 3D scanners significantly simplify the procedure of geometric modeling. 3D surfaces have become more widespread, leading to a great demand for noise removal with the expectation of the minimal disturbance of mesh geometry. We propose a novel two-step data-driven mesh denoising approach. The first step removes noise by learning normal variations from noisy models to their ground-truth counterparts. Unlike existing denoising methods, we present the second step to recover the mesh geometry lost in the first step. The second step understands the commonly used filters by learning the mapping from filtered models to their ground-truth counterparts. In addition, (1) to handle noise with large variations, we model normal estimation as a low-rank matrix recovery problem in similar-patch collaboration before the first-step learning; (2) to recover the real geometry of a denoised mesh, we reversely filter the denoised mesh to obtain more geometry cues before the second-step learning. The detailed quantitative and qualitative results on various data demonstrate that, our two-step learning algorithm competes favorably with the state-of-the-art methods in terms of mesh geometry preservation and noise-robustness.",
keywords = "Geometry-recovering, Low-rank matrix recovery, Mesh denoising, Normal variation learning, Reverse filter",
author = "Jun WANG and Jin HUANG and WANG, {Fu Lee} and Mingqiang WEI and Haoran XIE and Jing QIN",
year = "2019",
month = "9",
doi = "10.1016/j.cad.2019.05.027",
language = "English",
volume = "114",
pages = "133--142",
journal = "CAD Computer Aided Design",
issn = "0010-4485",
publisher = "Elsevier Ltd",

}

WANG, J, HUANG, J, WANG, FL, WEI, M, XIE, H & QIN, J 2019, 'Data-driven geometry-recovering mesh denoising', CAD Computer Aided Design, vol. 114, pp. 133-142. https://doi.org/10.1016/j.cad.2019.05.027

Data-driven geometry-recovering mesh denoising. / WANG, Jun; HUANG, Jin; WANG, Fu Lee; WEI, Mingqiang; XIE, Haoran; QIN, Jing.

In: CAD Computer Aided Design, Vol. 114, 09.2019, p. 133-142.

Research output: Journal PublicationsJournal Article (refereed)

TY - JOUR

T1 - Data-driven geometry-recovering mesh denoising

AU - WANG, Jun

AU - HUANG, Jin

AU - WANG, Fu Lee

AU - WEI, Mingqiang

AU - XIE, Haoran

AU - QIN, Jing

PY - 2019/9

Y1 - 2019/9

N2 - Depth cameras and 3D scanners significantly simplify the procedure of geometric modeling. 3D surfaces have become more widespread, leading to a great demand for noise removal with the expectation of the minimal disturbance of mesh geometry. We propose a novel two-step data-driven mesh denoising approach. The first step removes noise by learning normal variations from noisy models to their ground-truth counterparts. Unlike existing denoising methods, we present the second step to recover the mesh geometry lost in the first step. The second step understands the commonly used filters by learning the mapping from filtered models to their ground-truth counterparts. In addition, (1) to handle noise with large variations, we model normal estimation as a low-rank matrix recovery problem in similar-patch collaboration before the first-step learning; (2) to recover the real geometry of a denoised mesh, we reversely filter the denoised mesh to obtain more geometry cues before the second-step learning. The detailed quantitative and qualitative results on various data demonstrate that, our two-step learning algorithm competes favorably with the state-of-the-art methods in terms of mesh geometry preservation and noise-robustness.

AB - Depth cameras and 3D scanners significantly simplify the procedure of geometric modeling. 3D surfaces have become more widespread, leading to a great demand for noise removal with the expectation of the minimal disturbance of mesh geometry. We propose a novel two-step data-driven mesh denoising approach. The first step removes noise by learning normal variations from noisy models to their ground-truth counterparts. Unlike existing denoising methods, we present the second step to recover the mesh geometry lost in the first step. The second step understands the commonly used filters by learning the mapping from filtered models to their ground-truth counterparts. In addition, (1) to handle noise with large variations, we model normal estimation as a low-rank matrix recovery problem in similar-patch collaboration before the first-step learning; (2) to recover the real geometry of a denoised mesh, we reversely filter the denoised mesh to obtain more geometry cues before the second-step learning. The detailed quantitative and qualitative results on various data demonstrate that, our two-step learning algorithm competes favorably with the state-of-the-art methods in terms of mesh geometry preservation and noise-robustness.

KW - Geometry-recovering

KW - Low-rank matrix recovery

KW - Mesh denoising

KW - Normal variation learning

KW - Reverse filter

U2 - 10.1016/j.cad.2019.05.027

DO - 10.1016/j.cad.2019.05.027

M3 - Journal Article (refereed)

VL - 114

SP - 133

EP - 142

JO - CAD Computer Aided Design

JF - CAD Computer Aided Design

SN - 0010-4485

ER -