Background Subtraction with Dirichlet Process Mixture Models

Research output: Contribution to journalArticle

102 Citations (Scopus)
93 Downloads (Pure)

Abstract

Video analysis often begins with background subtraction. This problem is often approached in two steps - a background model followed by a regularisation scheme. A model of the background allows it to be distinguished on a per-pixel basis from the foreground, whilst the regularisation combines information from adjacent pixels. We present a new method based on Dirichlet process
Gaussian mixture models, which are used to estimate per-pixel background distributions. It is followed by probabilistic regularisation. Using a non-parametric Bayesian method allows per-pixel mode counts to be automatically inferred, avoiding over-/under- fitting. We also develop novel model learning algorithms for continuous update of the model in a principled fashion as the scene changes. These key advantages enable us to outperform the state-of-the-art alternatives on four benchmarks.
Original languageEnglish
Pages (from-to)670-683
Number of pages14
JournalIEEE Transactions on Pattern Analysis and Machine Intelligence
Volume36
Issue number4
DOIs
Publication statusPublished - 5 Dec 2013

Fingerprint

Background Subtraction
Dirichlet Process
Mixture Model
Process Model
Pixel
Regularization
Pixels
Video Analysis
Nonparametric Methods
Gaussian Mixture Model
Bayesian Methods
Model
Learning Algorithm
Count
Adjacent
Update
Learning algorithms
Benchmark
Alternatives
Estimate

Cite this

Background Subtraction with Dirichlet Process Mixture Models. / Fincham Haines, Tom; Xiang, Tao.

In: IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 36, No. 4, 05.12.2013, p. 670-683.

Research output: Contribution to journalArticle

@article{6f8ca61ae5084ad1850767d29ef2b80e,
title = "Background Subtraction with Dirichlet Process Mixture Models",
abstract = "Video analysis often begins with background subtraction. This problem is often approached in two steps - a background model followed by a regularisation scheme. A model of the background allows it to be distinguished on a per-pixel basis from the foreground, whilst the regularisation combines information from adjacent pixels. We present a new method based on Dirichlet processGaussian mixture models, which are used to estimate per-pixel background distributions. It is followed by probabilistic regularisation. Using a non-parametric Bayesian method allows per-pixel mode counts to be automatically inferred, avoiding over-/under- fitting. We also develop novel model learning algorithms for continuous update of the model in a principled fashion as the scene changes. These key advantages enable us to outperform the state-of-the-art alternatives on four benchmarks.",
author = "{Fincham Haines}, Tom and Tao Xiang",
year = "2013",
month = "12",
day = "5",
doi = "10.1109/TPAMI.2013.239",
language = "English",
volume = "36",
pages = "670--683",
journal = "IEEE Transactions on Pattern Analysis and Machine Intelligence",
issn = "0162-8828",
publisher = "IEEE",
number = "4",

}

TY - JOUR

T1 - Background Subtraction with Dirichlet Process Mixture Models

AU - Fincham Haines, Tom

AU - Xiang, Tao

PY - 2013/12/5

Y1 - 2013/12/5

N2 - Video analysis often begins with background subtraction. This problem is often approached in two steps - a background model followed by a regularisation scheme. A model of the background allows it to be distinguished on a per-pixel basis from the foreground, whilst the regularisation combines information from adjacent pixels. We present a new method based on Dirichlet processGaussian mixture models, which are used to estimate per-pixel background distributions. It is followed by probabilistic regularisation. Using a non-parametric Bayesian method allows per-pixel mode counts to be automatically inferred, avoiding over-/under- fitting. We also develop novel model learning algorithms for continuous update of the model in a principled fashion as the scene changes. These key advantages enable us to outperform the state-of-the-art alternatives on four benchmarks.

AB - Video analysis often begins with background subtraction. This problem is often approached in two steps - a background model followed by a regularisation scheme. A model of the background allows it to be distinguished on a per-pixel basis from the foreground, whilst the regularisation combines information from adjacent pixels. We present a new method based on Dirichlet processGaussian mixture models, which are used to estimate per-pixel background distributions. It is followed by probabilistic regularisation. Using a non-parametric Bayesian method allows per-pixel mode counts to be automatically inferred, avoiding over-/under- fitting. We also develop novel model learning algorithms for continuous update of the model in a principled fashion as the scene changes. These key advantages enable us to outperform the state-of-the-art alternatives on four benchmarks.

UR - https://doi.org/10.1109/TPAMI.2013.239

U2 - 10.1109/TPAMI.2013.239

DO - 10.1109/TPAMI.2013.239

M3 - Article

VL - 36

SP - 670

EP - 683

JO - IEEE Transactions on Pattern Analysis and Machine Intelligence

JF - IEEE Transactions on Pattern Analysis and Machine Intelligence

SN - 0162-8828

IS - 4

ER -