Joint NDT image restoration and segmentation using Gauss-Markov-Potts prior models and variational Bayesian...

Area of Science:

• Image processing
• Computer vision
• Statistical modeling

Background:

• Degraded images require simultaneous restoration and segmentation.
• Existing methods may not efficiently handle known point spread functions (PSF) and additive noise.

Purpose of the Study:

• To develop a Bayesian method for joint image restoration and segmentation.
• To model piecewise homogeneous images using nonhomogeneous Gauss-Markov fields with Potts region labels.

Main Methods:

• Utilizing a Bayesian estimation framework.
• Approximating the joint posterior law using variational Bayes techniques.
• Developing a practical joint restoration and segmentation algorithm.

Main Results:

• The proposed variational Bayes approximation enables a practical algorithm.
• Preliminary results demonstrate the effectiveness of the method.
• Comparison with MCMC Gibbs sampling is presented.

Conclusions:

• The developed prior models are suitable for images with finite homogeneous materials.
• The method is particularly relevant for nondestructive testing (NDT) applications.
• This approach offers an efficient solution for complex image restoration and segmentation tasks.