Assessment of sub-sampling schemes for compressive nano-FTIR imaging

TL;DR

This paper evaluates various practical sub-sampling schemes for nano-FTIR imaging, demonstrating that non-random methods at low sampling rates can achieve comparable results to random sampling, thus enabling faster data acquisition.

Contribution

It introduces and compares practical sub-sampling schemes for nano-FTIR imaging, showing they can replace random sampling for efficient, faster measurements.

Findings

Sub-sampling at 10% yields comparable results across schemes.

Non-random schemes like Lissajous are effective alternatives.

Random sub-sampling is not necessary for efficient data acquisition.

Abstract

Nano-FTIR imaging is a powerful scanning-based technique at nanometer spatial resolution which combines Fourier transform infrared spectroscopy (FTIR) and scattering-type scanning near-field optical microscopy (s-SNOM). However, recording large spatial areas with nano-FTIR is limited by long measurement times due to its sequential data acquisition. Several mathematical approaches have been proposed to tackle this problem. All of them have in common that only a small fraction of randomly chosen measurements is required. However, choosing the fraction of measurements in a random fashion poses practical challenges for scanning procedures and does not lead to time savings as large as desired. We consider different, practically relevant sub-sampling schemes assuring a faster acquisition. It is demonstrated that results for almost all considered sub-sampling schemes, namely original Lissajous, triangle Lissajous, and random reflection sub-sampling, at a sub-sampling rate of 10%, are comparable to results when using a random sub-sampling of 10%. This implies that random sub-sampling is not required for efficient data acquisition.