Speckle suppression in SAR images using the 2-D GARCH model

Area of Science:

• Remote Sensing
• Signal Processing
• Image Analysis

Background:

• Speckle noise significantly degrades Synthetic Aperture Radar (SAR) image quality.
• Existing speckle suppression methods often compromise structural and textural information.
• Understanding the statistical properties of SAR image wavelet coefficients is crucial for effective noise reduction.

Purpose of the Study:

• To develop a novel Bayesian-based speckle suppression method for SAR images.
• To preserve structural features and textural information during noise reduction.
• To improve the performance of speckle suppression compared to existing techniques.

Main Methods:

• Logarithmic transform of SAR images followed by multiscale wavelet domain analysis.
• Modeling wavelet coefficients using the 2-D Generalized Autoregressive Conditional Heteroskedasticity (GARCH) model to capture non-Gaussian statistics and dependencies.
• Employing a Maximum A Posteriori (MAP) estimator for clean image wavelet coefficient estimation.

Main Results:

• The 2-D GARCH model accurately describes the non-Gaussian statistics of SAR image wavelet coefficients.
• The proposed Bayesian method effectively suppresses speckle while preserving image structural and textural details.
• Performance improvements were verified on both synthetic and actual SAR images.

Conclusions:

• The novel Bayesian speckle suppression method offers superior performance in SAR image processing.
• The use of the 2-D GARCH model in the wavelet domain is key to preserving image fidelity.
• This approach represents a significant advancement in SAR image denoising.