Self-similar texture modeling using FARIMA processes with applications to satellite images

Area of Science:

• Remote Sensing
• Image Processing
• Statistical Modeling

Background:

• Developing accurate texture models for Synthetic Aperture Radar (SAR) sea surface images is crucial for ocean surveillance.
• Existing models like Moving-Average (MA), Autoregressive (AR), and Fractionally Differenced (FD) have limitations in capturing complex spatial dependencies.

Discussion:

• A novel texture model employing a two-dimensional (2-D) fractionally integrated autoregressive-moving average (FARIMA) process with a non-Gaussian driving sequence is proposed for SAR sea surface imagery.
• The FARIMA model effectively captures both long-range and short-range spatial dependencies using a parsimonious set of parameters.
• An efficient spectral fitting procedure is presented for estimating the FARIMA model parameters.

Key Insights:

• The proposed 2-D FARIMA model provides a more accurate description of SAR sea surface images compared to conventional MA, AR, and FD models.
• The model's ability to capture long-range spatial dependence is a key advantage for analyzing complex sea surface textures.
• Validation using RADARSAT ocean surveillance data demonstrates the practical applicability and superior performance of the FARIMA approach.

Outlook:

• Further research could explore extensions of the FARIMA model for different types of SAR imagery or environmental conditions.
• Investigating the application of this model in advanced maritime surveillance systems and sea state monitoring is recommended.
• Potential for integrating this texture model into machine learning pipelines for enhanced feature extraction in remote sensing.