The Gamma Generalized Normal Distribution: A Descriptor of SAR Imagery

TL;DR

This paper introduces the gamma generalized normal (GGN) distribution, a new four-parameter model for SAR imagery that combines gamma and generalized normal distributions, with promising performance over existing models.

Contribution

The paper develops and characterizes the GGN distribution, providing methods for estimation and demonstrating its effectiveness on SAR data compared to existing models.

Findings

GGN can outperform BGN in modeling SAR imagery

Mathematical properties of GGN are thoroughly characterized

Performance evaluated on synthetic and real SAR data

Abstract

We propose a new four-parameter distribution for modeling synthetic aperture radar (SAR) imagery named the gamma generalized normal (GGN) by combining the gamma and generalized normal distributions. A mathematical characterization of the new distribution is provided by identifying the limit behavior and by calculating the density and moment expansions. The GGN model performance is evaluated on both synthetic and actual data and, for that, maximum likelihood estimation and random number generation are discussed. The proposed distribution is compared with the beta generalized normal distribution (BGN), which has already shown to appropriately represent SAR imagery. The performance of these two distributions are measured by means of statistics which provide evidence that the GGN can outperform the BGN distribution in some contexts.