Signals and images foreground/background joint estimation and separation

Boujemaa Ait-El-Fquih*, Ali Mohammad-Djafari

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

This paper is devoted to a foreground/background joint estimation and separation problem. We first observe that this problem is modeled by a conditionally linear and Gaussian hidden Markov chain (CLGHMC). We next propose a filtering algorithm in the general non-linear and non Gaussian conditionally hidden Markov chain (CHMC), allowing the propagation of the filtering densities associated to the foreground and the background. We then focus on the particular case of our CLGHMC in which these filtering densities are weighted sums of Gaussian distributions; the parameters of each Gaussian are computed by using the Kalman filter algorithm, while the weights are computed by using the particle filter algorithm. We finally perform some simulations to highlight the interest of our method in both signals and images foreground/backgound joint estimation and separation.

Original languageEnglish (US)
Title of host publicationBayesian Inference and Maximum Entropy Methods in Science and Engineering - Proc. of the 30th Intl. Workshop on Bayesian Inference and Maximum Entropy Methods in Science and Engineering, MaxEnt 2010
Pages266-273
Number of pages8
Volume1305
DOIs
StatePublished - 2010
Externally publishedYes
Event30th International Workshop on Bayesian Inference and Maximum Entropy Methods in Science and Engineering, MaxEnt 2010 - Chamonix, France
Duration: Jul 4 2010Jul 9 2010

Other

Other30th International Workshop on Bayesian Inference and Maximum Entropy Methods in Science and Engineering, MaxEnt 2010
CountryFrance
CityChamonix
Period07/4/1007/9/10

Keywords

  • Conditionally hidden Markov chain models
  • Foreground/background joint estimation and separation
  • Kalman filter
  • Particle filter

ASJC Scopus subject areas

  • Physics and Astronomy(all)

Fingerprint Dive into the research topics of 'Signals and images foreground/background joint estimation and separation'. Together they form a unique fingerprint.

Cite this