Clustering of Mueller matrix images for skeletonized structure detection

Area of Science:

• Image analysis
• Computational biology
• Biophysics

Background:

• Polarization-encoded images offer rich information for material characterization.
• Clustering techniques are crucial for analyzing complex image data.
• Previous methods for polarization image clustering require refinement.

Purpose of the Study:

• To extend and refine clustering methods for polarization-encoded images.
• To apply Markovian Bayesian inference for analyzing multidimensional parametric images.
• To evaluate the effectiveness of Hidden Markov Models for Mueller images.

Main Methods:

• Clustering scheme based on Markovian Bayesian inference.
• Application of Hidden Markov Chains Model (HMCM) and Hidden Hierarchical Markovian Model (HHMM).
• Analysis of multidimensional parametric images, specifically Mueller images.

Main Results:

• HMCM and HHMM effectively handle Mueller images.
• The proposed methods yield excellent results for biological tissues, such as vegetal leaves.
• Principal Component Analysis (PCA) for dimensionality reduction is ineffective for Mueller matrix images.

Conclusions:

• Markovian Bayesian inference provides a robust framework for clustering polarization-encoded images.
• Hidden Markov Models are highly effective for analyzing Mueller images of biological tissues.
• Dimensionality reduction techniques like PCA are not suitable for Mueller matrix image analysis in this context.