Enhancing RABASAR for Multi-Temporal SAR Image Despeckling through Directional Filtering and Wavelet Transform

Area of Science:

• Remote Sensing
• Image Processing
• Geospatial Analysis

Background:

• Speckle noise significantly degrades synthetic aperture radar (SAR) image interpretability.
• Existing despeckling methods for single-temporal SAR images are well-developed, but multi-temporal SAR image despeckling remains a challenge.
• Limitations exist in current frameworks regarding 'superimage' acquisition and ratio image generation for multi-temporal SAR data.

Purpose of the Study:

• To propose an enhanced framework for despeckling multi-temporal SAR images.
• To address limitations in 'superimage' acquisition and ratio image generation within existing despeckling frameworks.
• To improve the interpretability and information fusion capabilities of multi-temporal SAR imagery.

Main Methods:

• Developed a direction-based segmentation approach for multi-temporal SAR non-local means filtering (DSMT-NLM) to obtain the 'superimage'.
• Extended the non-local means (NLM) algorithm to multi-temporal SAR images using directional segmentation.
• Employed a weighted averaging method based on wavelet transform (WAMWT) for enhanced ratio image generation.

Main Results:

• The proposed DSMT-NLM method effectively removes speckle noise from multi-temporal SAR images.
• The framework reduces the generation of false details compared to existing methods (RABASAR, Frost, NLM).
• Successful fusion of multi-temporal information was achieved, enhancing overall image utility.

Conclusions:

• The enhanced framework significantly outperforms RABASAR, Frost, and NLM in multi-temporal SAR image despeckling.
• The proposed DSMT-NLM and WAMWT methods provide a robust solution for noise reduction and information fusion.
• This approach enhances the interpretability and application potential of multi-temporal SAR data.