Classification of complex Wishart matrices with a diffusion-reaction system guided by stochastic distances

Area of Science:

• Remote Sensing
• Image Processing
• Statistical Modeling

Background:

• Polarimetric synthetic aperture radar (PolSAR) imagery provides rich information for land cover classification.
• Accurate classification of PolSAR data is crucial for various applications, but is often challenged by speckle noise.

Purpose of the Study:

• To develop an advanced classification method for PolSAR imagery.
• To improve classification accuracy and reduce speckle noise in PolSAR data.

Main Methods:

• A new classification method based on stochastic distances in the space of complex Wishart distributed random matrices.
• An iterative classification approach using weighted stochastic distances to maximize class discrimination.
• Integration of a diffusion-reaction partial differential system for image smoothing and prototype refinement.

Main Results:

• The proposed method effectively classifies PolSAR imagery.
• Demonstrated significant speckle reduction capabilities inherited from diffusion processes.
• Performance validated on both synthetic and real-world PolSAR datasets.

Conclusions:

• The novel stochastic distance-based method offers a robust solution for PolSAR image classification.
• The diffusion-reaction system enhances classification by smoothing and guiding pixel values.
• The method shows strong potential for improving PolSAR data analysis and applications.