Multi-spectral image shift-estimation error calculations using simulated phenomenology.

Area of Science:

• Remote Sensing
• Image Processing
• Computational Imaging

Background:

• Multi-spectral image registration is crucial for applications like image fusion.
• Phenomenological differences across spectral bands introduce significant errors in registration parameter estimation.
• Quantifying these errors with archival data is challenging due to inherent limitations.

Purpose of the Study:

• To develop a method for creating high-fidelity synthetic multi-spectral imagery.
• To quantify algorithmic and fundamental errors in multi-spectral image registration.
• To establish credible requirements for multi-spectral imaging systems.

Main Methods:

• Utilized commercial graphics software and optical property measurements to generate synthetic multi-spectral data.
• Developed techniques for error quantification in area-based and feature-based registration algorithms.
• Employed synthetic datasets to calculate Fisher information and sample errors for shift estimation.

Main Results:

• Demonstrated the creation of synthetic multi-spectral imagery with accurate emissive and reflective properties.
• Successfully quantified registration errors using both area- and feature-based methods on synthetic data.
• Showcased the ability to determine Fisher information and sample errors for inter-band image shifts.

Conclusions:

• Synthetic data generation offers a flexible and robust approach to characterizing multi-spectral image registration errors.
• This method enables error quantification across various image brightness, sensor parameters, and phenomenological differences.
• The findings support the development of more accurate and reliable multi-spectral imaging systems.