Employing a Mixture of Rayleigh and Inverse Gaussian Distributions as SAR Clutter Texture Model in MAP Estimation for Efficient Speckle Suppression

Employing a Mixture of Rayleigh and Inverse Gaussian Distributions as SAR Clutter Texture Model in MAP Estimation for Efficient Speckle Suppression

This article proposes a speckle suppression technique based on a binary mixture of the Rayleigh and the reciprocal of Gaussian (RIG) distributions. This model suitably characterizes the texture of synthetic aperture radar (SAR) return from regions with varying degrees of roughness and captures the m...

Full description

Saved in:
Bibliographic Details
Main Authors: Dheeren Ku Mahapatra, Saurav Gupta, Biswajit Jena, Ravi Prakash Dwivedi
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Access
Subjects:
SAR clutter amplitude
speckle
Rayleigh-inverse Gaussian mixture model
expectation-maximization
Online Access:https://ieeexplore.ieee.org/document/11002505/
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This article proposes a speckle suppression technique based on a binary mixture of the Rayleigh and the reciprocal of Gaussian (RIG) distributions. This model suitably characterizes the texture of synthetic aperture radar (SAR) return from regions with varying degrees of roughness and captures the multimodal behavior observed in extremely heterogeneous SAR clutter. We estimate the RIG mixture model parameters by maximum likelihood (ML) with the expectation maximization (EM) algorithm. We also obtain the Cr&#x00E1;mer-Rao Bounds (CRBs) for these estimators. Finally, we propose a maximum-a-posteriori (MAP) estimator for efficient despeckling by utilizing the RIG model as a prior distribution for the texture component. The accuracy of RIG-MAP estimation for texture is performed on single-look clutter data from actual sensors and multilook simulated clutter data. Qualitative and quantitative results on despeckling illustrate the effectiveness of the proposed MAP estimator in suppressing speckle while preserving mean, textural information, fine details, etc. Furthermore, the RIG-MAP estimator achieves superior performance compared to MMSE (minimum mean square error) - based Lee filter, Kuan filter, and MAP-based (such as <inline-formula> <tex-math notation="LaTeX">$\beta $ </tex-math></inline-formula>-MAP, <inline-formula> <tex-math notation="LaTeX">$\Gamma $ </tex-math></inline-formula>-MAP, CR-MAP, <inline-formula> <tex-math notation="LaTeX">$\mathcal {G}^{0}$ </tex-math></inline-formula>-MAP) estimators.
ISSN:2169-3536

Similar Items