Removal of Spectro-Polarimetric Fringes by 2D Pattern Recognition

TL;DR

This paper introduces a 2D pattern recognition method using Principal Component Analysis to effectively remove polarized fringes from spectro-polarimetric data, improving data quality for scientific analysis.

Contribution

The novel approach applies 2D PCA trained on spectro-polarimetric maps to identify and eliminate fringe structures, offering a new method for data cleaning.

Findings

Effective fringe removal demonstrated on spectro-polarimetric data

Potential to improve calibration and data analysis workflows

Suggests changes in data acquisition planning

Abstract

We present a pattern-recognition based approach to the problem of removal of polarized fringes from spectro-polarimetric data. We demonstrate that 2D Principal Component Analysis can be trained on a given spectro-polarimetric map in order to identify and isolate fringe structures from the spectra. This allows us in principle to reconstruct the data without the fringe component, providing an effective and clean solution to the problem. The results presented in this paper point in the direction of revising the way that science and calibration data should be planned for a typical spectro-polarimetric observing run.