Compound truncated Poisson gamma distribution for understanding multimodal SAR intensities.

Area of Science:

• Probability theory and statistical modeling
• Data analysis and signal processing

Background:

• Mixture models offer flexibility but often involve complex, high-dimensional parameter spaces.
• Existing models can be computationally intensive for inference, posing challenges in practical applications.

Purpose of the Study:

• To propose a novel, parsimonious probability model for multimodal data.
• To introduce the compound truncated Poisson gamma (CTrPGa) distribution with only three parameters.
• To explore the theoretical properties and estimation methods for the CTrPGa distribution.

Main Methods:

• Derivation of key properties: hazard, characteristic, cumulative functions, and ordinary moments.
• Development of estimation techniques: moment estimation, maximum likelihood estimation (Expectation-Maximization), and empirical characteristic function.
• Simulation studies to compare estimation method performance.
• Application to real synthetic aperture radar (SAR) imagery.

Main Results:

• The CTrPGa distribution offers a flexible yet parsimonious approach to modeling multimodal data.
• Moment estimation is simplified, requiring the solution of a single nonlinear equation.
• Simulation analysis demonstrates the comparative performance of the estimation methods.
• Real SAR imagery experiments indicate the CTrPGa model's suitability for pre-processing.

Conclusions:

• The compound truncated Poisson gamma (CTrPGa) distribution is a viable and efficient alternative for modeling complex data.
• The proposed model and its estimation methods are applicable to real-world problems, particularly in SAR image analysis.
• The CTrPGa distribution can be a valuable tool in the pre-processing pipeline for SAR imagery.