Multitemporal SAR Image Despeckling Based on a Scattering Covariance Matrix of Image Patch

Area of Science:

• Remote Sensing
• Signal Processing

Background:

• Synthetic Aperture Radar (SAR) images are prone to speckle noise, which degrades image quality and hinders analysis.
• Multitemporal SAR data offers enhanced information but requires robust methods to handle speckle across different time phases.

Purpose of the Study:

• To develop an advanced despeckling method for multitemporal SAR images.
• To effectively utilize both amplitude and phase information within image patches for noise reduction.

Main Methods:

• A novel filtering framework processing image patches using scattering covariance matrices.
• Incorporation of nonlocal weighted averaging, adaptive temporal filtering, and guided filtering.
• An aggregation step to resolve multiple estimations for each pixel.

Main Results:

• The proposed method demonstrates significant speckle suppression in simulated and real multitemporal SAR datasets.
• The technique successfully retains fine details and structural information within the images.
• Comparative analysis shows superior performance over existing single-temporal and multitemporal SAR despeckling methods.

Conclusions:

• The developed method offers a powerful solution for despeckling multitemporal SAR images.
• It provides a good balance between noise reduction and detail preservation.
• The approach is effective for various SAR imaging applications requiring high-quality data.