Mean field annealing using compound Gauss-Markov random fields for edge detection and image estimation

Area of Science:

• Image processing
• Computer vision
• Computational statistics

Background:

• Nonstationary images are susceptible to additive Gaussian noise, complicating edge detection and image estimation.
• Traditional methods often require computationally intensive stochastic relaxation algorithms due to non-convex posterior probability functions.

Purpose of the Study:

• To propose a deterministic relaxation method for edge detection and image estimation in nonstationary images.
• To utilize a compound Gauss-Markov random field (CGMRF) model for image representation.
• To address the limitations of stochastic algorithms in maximum a posteriori (MAP) estimation.

Main Methods:

• A deterministic mean field annealing algorithm is developed for image estimation and edge detection.
• Iterative equations are derived for mean values of intensity and line processes (horizontal and vertical).
• The CGMRF model is employed to represent the noise-free image structure.

Main Results:

• The proposed method provides effective edge detection and image estimation results on noisy images.
• It offers a computationally efficient alternative to stochastic relaxation methods.
• The study analyzes the interaction between intensity and line processes.

Conclusions:

• The deterministic mean field annealing approach with a CGMRF model is a viable and efficient technique for image estimation and edge detection.
• This method demonstrates robust performance in handling nonstationary images corrupted by Gaussian noise.
• The findings contribute to advancements in image restoration and feature extraction techniques.