Probabilistic image-based rendering with Gaussian mixture model

Wenfeng Li, Baoxin Li

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

2 Citations (Scopus)

Abstract

One major challenge in traditional image-based rendering is 3D scene reconstruction by estimating accurate dense depth map, which suffers from the ambiguities in textureless or periodically textured regions. Alternatively, statistical methods may be used to estimate a most likely color for each pixel for photorealistic rendering from multiple views of the same scene. Such statistical methods normally require a relatively large number of input images to achieve reasonable quality for the synthesized image, if the estimation is purely nonparametric. In this paper, based on some reasonable assumptions on the configuration of the multiple views, we propose to use a two-component Gaussian mixture model for the appearance of a given pixel in all the views so that both the problem of occlusion and the problem of noise can be considered simultaneously. Then we use the Expectation-Maximization algorithm to estimate the model parameters. The virtual pixel is given as a maximum likelihood estimate for one of the mixture components. Experiments shows that reasonable performance can be obtained even with only a few input images.

Original languageEnglish (US)
Title of host publicationProceedings - International Conference on Pattern Recognition
Pages179-182
Number of pages4
Volume1
DOIs
StatePublished - 2006
Event18th International Conference on Pattern Recognition, ICPR 2006 - Hong Kong, China
Duration: Aug 20 2006Aug 24 2006

Other

Other18th International Conference on Pattern Recognition, ICPR 2006
CountryChina
CityHong Kong
Period8/20/068/24/06

Fingerprint

Pixels
Statistical methods
Maximum likelihood
Color
Experiments

ASJC Scopus subject areas

  • Electrical and Electronic Engineering
  • Computer Vision and Pattern Recognition
  • Hardware and Architecture

Cite this

Li, W., & Li, B. (2006). Probabilistic image-based rendering with Gaussian mixture model. In Proceedings - International Conference on Pattern Recognition (Vol. 1, pp. 179-182). [1698862] https://doi.org/10.1109/ICPR.2006.945

Probabilistic image-based rendering with Gaussian mixture model. / Li, Wenfeng; Li, Baoxin.

Proceedings - International Conference on Pattern Recognition. Vol. 1 2006. p. 179-182 1698862.

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

Li, W & Li, B 2006, Probabilistic image-based rendering with Gaussian mixture model. in Proceedings - International Conference on Pattern Recognition. vol. 1, 1698862, pp. 179-182, 18th International Conference on Pattern Recognition, ICPR 2006, Hong Kong, China, 8/20/06. https://doi.org/10.1109/ICPR.2006.945
Li W, Li B. Probabilistic image-based rendering with Gaussian mixture model. In Proceedings - International Conference on Pattern Recognition. Vol. 1. 2006. p. 179-182. 1698862 https://doi.org/10.1109/ICPR.2006.945
Li, Wenfeng ; Li, Baoxin. / Probabilistic image-based rendering with Gaussian mixture model. Proceedings - International Conference on Pattern Recognition. Vol. 1 2006. pp. 179-182
@inproceedings{45b6f1116a5147ce8ef187d01712c065,
title = "Probabilistic image-based rendering with Gaussian mixture model",
abstract = "One major challenge in traditional image-based rendering is 3D scene reconstruction by estimating accurate dense depth map, which suffers from the ambiguities in textureless or periodically textured regions. Alternatively, statistical methods may be used to estimate a most likely color for each pixel for photorealistic rendering from multiple views of the same scene. Such statistical methods normally require a relatively large number of input images to achieve reasonable quality for the synthesized image, if the estimation is purely nonparametric. In this paper, based on some reasonable assumptions on the configuration of the multiple views, we propose to use a two-component Gaussian mixture model for the appearance of a given pixel in all the views so that both the problem of occlusion and the problem of noise can be considered simultaneously. Then we use the Expectation-Maximization algorithm to estimate the model parameters. The virtual pixel is given as a maximum likelihood estimate for one of the mixture components. Experiments shows that reasonable performance can be obtained even with only a few input images.",
author = "Wenfeng Li and Baoxin Li",
year = "2006",
doi = "10.1109/ICPR.2006.945",
language = "English (US)",
isbn = "0769525210",
volume = "1",
pages = "179--182",
booktitle = "Proceedings - International Conference on Pattern Recognition",

}

TY - GEN

T1 - Probabilistic image-based rendering with Gaussian mixture model

AU - Li, Wenfeng

AU - Li, Baoxin

PY - 2006

Y1 - 2006

N2 - One major challenge in traditional image-based rendering is 3D scene reconstruction by estimating accurate dense depth map, which suffers from the ambiguities in textureless or periodically textured regions. Alternatively, statistical methods may be used to estimate a most likely color for each pixel for photorealistic rendering from multiple views of the same scene. Such statistical methods normally require a relatively large number of input images to achieve reasonable quality for the synthesized image, if the estimation is purely nonparametric. In this paper, based on some reasonable assumptions on the configuration of the multiple views, we propose to use a two-component Gaussian mixture model for the appearance of a given pixel in all the views so that both the problem of occlusion and the problem of noise can be considered simultaneously. Then we use the Expectation-Maximization algorithm to estimate the model parameters. The virtual pixel is given as a maximum likelihood estimate for one of the mixture components. Experiments shows that reasonable performance can be obtained even with only a few input images.

AB - One major challenge in traditional image-based rendering is 3D scene reconstruction by estimating accurate dense depth map, which suffers from the ambiguities in textureless or periodically textured regions. Alternatively, statistical methods may be used to estimate a most likely color for each pixel for photorealistic rendering from multiple views of the same scene. Such statistical methods normally require a relatively large number of input images to achieve reasonable quality for the synthesized image, if the estimation is purely nonparametric. In this paper, based on some reasonable assumptions on the configuration of the multiple views, we propose to use a two-component Gaussian mixture model for the appearance of a given pixel in all the views so that both the problem of occlusion and the problem of noise can be considered simultaneously. Then we use the Expectation-Maximization algorithm to estimate the model parameters. The virtual pixel is given as a maximum likelihood estimate for one of the mixture components. Experiments shows that reasonable performance can be obtained even with only a few input images.

UR - http://www.scopus.com/inward/record.url?scp=34047204771&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=34047204771&partnerID=8YFLogxK

U2 - 10.1109/ICPR.2006.945

DO - 10.1109/ICPR.2006.945

M3 - Conference contribution

SN - 0769525210

SN - 9780769525211

VL - 1

SP - 179

EP - 182

BT - Proceedings - International Conference on Pattern Recognition

ER -