[Paper Review] High-resolution wide-field magnetic imaging with sparse sampling using nitrogen-vacancy centers

Keqing Liu, Jiazhao Tian+8 more|arXiv (Cornell University)|Jan 31, 2026

Diamond and Carbon-based Materials Research0 citations

TL;DR

Proposes a sparse-sampling framework (MABE) to reconstruct 10^4-pixel wide-field NV-diamond magnetic-field images from only 25 measurements, with ~SSIM > 0.999 and enhanced sensitivity via phase-modulated dynamical decoupling pulses.

ABSTRACT

Nitrogen-vacancy (NV) centers in diamond enable quantitative magnetic imaging, yet practical implementations must balance spatial resolution against acquisition time (and thus per-pixel sensitivity). Single-NV scanning magnetometry achieves genuine nanoscale resolution, nonetheless requires typically a slow pixel-by-pixel acquisition. Meanwhile, wide-field NV-ensemble microscopy provides parallel readout over a large field of view, however is jointly limited by the optical diffraction limit and the sensor-sample standoff. Here, we present a sparse-sampling strategy for reconstructing high-resolution wide-field images from only a small number of measurements. Using simulated NV-ensemble detection of ac magnetic fields, we show that a mean-adjusted Bayesian estimation (MABE) framework can reconstruct 10000-pixel images from only 25 sampling points, achieving SSIM values exceeding 0.999 for representative smooth field distributions, while optimized dynamical-decoupling pulse sequences yield an approximately twofold improvement in magnetic-field sensitivity. The method further clarifies how sampling patterns and sampling density affect reconstruction accuracy and suggests a route toward faster and more scalable magnetic-imaging architectures that may extend to point-scanning NV sensors and other magnetometry platforms, such as SQUIDs, Hall probes, and magnetic tunnel junctions.

Motivation & Objective

• Motivate fast, wide-field magnetic imaging with NV centers by balancing spatial resolution and acquisition time.
• Demonstrate that sparse sampling plus Bayesian reconstruction can recover high-resolution magnetic-field maps.
• Show how optimized dynamical decoupling pulses improve per-pixel sensitivity.
• Provide guidance on sampling patterns and densities for accurate reconstruction.
• Suggest applicability to other magnetometry platforms beyond NV centers.

Proposed method

• Simulate NV ensemble detection of ac magnetic fields under realistic noise, with a 400 μs total evolution time and XY-8 sequences.
• Apply mean-adjusted Bayesian estimation (MABE) to reconstruct 10^4-pixel images from 25 sampling points.
• Introduce a reference-based mean-adjustment (proportional calibration) to correct bias in measurements.
• Optimize phase-modulated (PM) control pulses to maximize gate fidelity under inhomogeneous broadening and noise.
• Quantify imaging performance with SSIM and other metrics across distributions with 1, 2, and 3 extrema.

Experimental results

Research questions

• RQ1 Can high-resolution wide-field magnetic-field images be faithfully reconstructed from a sparse set of measurements using NV ensembles?
• RQ2 How do sampling density and sampling pattern affect reconstruction quality under realistic NV noise?
• RQ3 What improvements in magnetic-field sensitivity can be achieved with PM-pulse optimized dynamical decoupling?
• RQ4 Is the MABE framework training-free and scalable to larger fields of view or other magnetometry platforms?

Key findings

• A 10^4-pixel magnetic-field image can be reconstructed from 25 sampling points with SSIM > 0.999 for representative smooth distributions.
• PM-pulse optimization increases Pauli-X and Pauli-Y gate fidelities from ~0.43 to ~0.68, improving spin-contrast and sensitivity.
• Optimized sensing sequences yield about a twofold improvement in magnetic-field sensitivity (e.g., from 0.94 to 0.45 nT/√Hz in one case).
• MABE reconstructions with reference-based mean-adjustment achieve substantial metric improvements (MAE and RMSE drop to ~10^-4; PSNR ~30 dB; R^2 ~0.98–0.99; SSIM ~1.0).
• Grid-based sampling strategies perform best among tested strategies for sparse reconstruction.
• Increasing sampling points from 25 to 100 yields diminishing returns in SSIM (>0.9999) despite a fourfold increase in samples.