Polarimetric SAR Speckle Filtering Using a Nonlocal Weighted LMMSE Filter

Area of Science:

• Remote Sensing
• Signal Processing

Background:

• Speckle noise significantly degrades Polarimetric Synthetic Aperture Radar (PolSAR) data quality.
• Effective despeckling is crucial for accurate analysis and interpretation of PolSAR imagery.

Purpose of the Study:

• To propose a novel nonlocal weighted linear minimum mean-squared error (NWLMMSE) filter for PolSAR speckle reduction.
• To enhance the preservation of point targets, edges, and scattering mechanisms during the filtering process.

Main Methods:

• Development of the NWLMMSE estimator by integrating nonlocal means into the LMMSE framework.
• Implementation of a preclassification step to group pixels based on scattering mechanisms.
• Evaluation using simulated and real-world PolSAR datasets.

Main Results:

• The NWLMMSE filter demonstrates significant speckle reduction capabilities.
• The proposed method effectively preserves strong point targets and image edges.
• Polarimetric scattering mechanisms are well-maintained after filtering.

Conclusions:

• The NWLMMSE filter offers a robust solution for PolSAR despeckling.
• This technique improves the quality of PolSAR data for various applications.
• The filter balances noise reduction with the preservation of essential image features.