Bayesian image classification using Markov random fields

Berthod, Marc-Antoine; Kató, Zoltán; Yu, Shan; Zerubia, Josiane

doi:10.1016/0262-8856(95)01072-6

Cited by 146 publications

(107 citation statements)

References 10 publications

Supporting

Mentioning

105

Contrasting

Unclassified

Order By: Relevance

“…The most promising methods within this category are based on Markov random fields (MRF), a multidimensional extension of Markov chains [Kindermann and Snell, 1980;Chellappa and Jain, 1993]. These methods are usually insensitive to local image noise and produce reasonable results, but may require a priori knowledge for construction of an appropriate image model, a supervised learning process or manual initialization [e.g., Berthod et al, 1996;Szirányi et al, 2000;Li, 2001].…”

Section: Overview Of Segmentation Techniquesmentioning

confidence: 99%

“…Unfortunately, as with any deformable surface or region growing approach, this method heavily depends on initial seeding (thresholding) and the implementation of the speed function (sensitivity to between-class intensity contrast and local gradient values) is governing contour evolution. [44] MRF-Berthod is an algorithm for supervised Bayesian segmentation developed by Berthod et al [1996]. We adopted a C ++ code written by Csaba Gradwohl and Zoltan Kato (University of Szeged, Hungary) to implement this algorithm.…”

Section: Locally Adaptive Segmentationmentioning

confidence: 99%

See 1 more Smart Citation

Segmentation of X‐ray computed tomography images of porous materials: A crucial step for characterization and quantitative analysis of pore structures

Iassonov

Gebrenegus

Tuller

2009

Water Resources Research

497

332

View full text Add to dashboard Cite

[1] Nondestructive imaging methods such as X-ray computed tomography (CT) yield high-resolution, three-dimensional representations of pore space and fluid distribution within porous materials. Steadily increasing computational capabilities and easier access to X-ray CT facilities have contributed to a recent surge in microporous media research with objectives ranging from theoretical aspects of fluid and interfacial dynamics at the pore scale to practical applications such as dense nonaqueous phase liquid transport and dissolution. In recent years, significant efforts and resources have been devoted to improve CT technology, microscale analysis, and fluid dynamics simulations. However, the development of adequate image segmentation methods for conversion of gray scale CT volumes into a discrete form that permits quantitative characterization of pore space features and subsequent modeling of liquid distribution and flow processes seems to lag. In this paper we investigated the applicability of various thresholding and locally adaptive segmentation techniques for industrial and synchrotron X-ray CT images of natural and artificial porous media. A comparison between directly measured and image-derived porosities clearly demonstrates that the application of different segmentation methods as well as associated operator biases yield vastly differing results. This illustrates the importance of the segmentation step for quantitative pore space analysis and fluid dynamics modeling. Only a few of the tested methods showed promise for both industrial and synchrotron tomography. Utilization of local image information such as spatial correlation as well as the application of locally adaptive techniques yielded significantly better results.Citation: Iassonov, P., T. Gebrenegus, and M. Tuller (2009), Segmentation of X-ray computed tomography images of porous materials: A crucial step for characterization and quantitative analysis of pore structures, Water Resour. Res., 45, W09415,

show abstract

Section: Overview Of Segmentation Techniquesmentioning

confidence: 99%

Section: Locally Adaptive Segmentationmentioning

confidence: 99%

Segmentation of X‐ray computed tomography images of porous materials: A crucial step for characterization and quantitative analysis of pore structures

Iassonov

Gebrenegus

Tuller

2009

Water Resources Research

497

332

View full text Add to dashboard Cite

show abstract

“…Due to the fine spatial detail contained in the VFSR imagery, the parameter γ controlling the level of smoothness was set as 0.7 to achieve an increasing level of smoothness in terms of the MRF. The simulated annealing optimization using a Gibbs sampler [55] was employed in MLP-MRF to maximize the posterior probability through iteration.…”

Section: B Model Architectures and Parameter Settingsmentioning

confidence: 99%

VPRS-Based Regional Decision Fusion of CNN and MRF Classifications for Very Fine Resolution Remotely Sensed Images

Zhang

Sargent

Pan

et al. 2018

IEEE Trans. Geosci. Remote Sensing

View full text Add to dashboard Cite

show abstract

“…We can cite here some models widely used like the structural approach by regions growth [25], the stochastic approaches [11,12,16,17,21] and the variational approaches which are based on various strategies like level set formulations, the Mumford-Shah functional, active contours and geodesic active contours methods or wavelet transforms [7-9, 23, 24, 28, 29, 35].…”

Section: Introductionmentioning

confidence: 99%

A topological asymptotic analysis for the regularized grey-level image classification problem

2007

View full text Add to dashboard Cite

Abstract. The aim of this article is to propose a new method for the grey-level image classification problem. We first present the classical variational approach without and with a regularization term in order to smooth the contours of the classified image. Then we present the general topological asymptotic analysis, and we finally introduce its application to the grey-level image classification problem.

show abstract

Bayesian image classification using Markov random fields

Cited by 146 publications

References 10 publications

Segmentation of X‐ray computed tomography images of porous materials: A crucial step for characterization and quantitative analysis of pore structures

Segmentation of X‐ray computed tomography images of porous materials: A crucial step for characterization and quantitative analysis of pore structures

VPRS-Based Regional Decision Fusion of CNN and MRF Classifications for Very Fine Resolution Remotely Sensed Images

A topological asymptotic analysis for the regularized grey-level image classification problem

Contact Info

Product

Resources

About