Camera-Based Wide-Field Quantum Noise Spectroscopy

• Camera-based wide-field quantum noise spectroscopy is a measurement paradigm that uses camera sensors to parallelize quantum noise, correlations, and decoherence as image-forming observables.
• It integrates diverse platforms such as twin-beam, SPAD arrays, and NV-center fluorescence to extract spatially resolved quantum noise characteristics across large fields-of-view.
• The method employs estimator-driven models and advanced reconstruction techniques to achieve sub-shot-noise performance and improved sensitivity in imaging and quantum sensing.

Searching arXiv for papers on camera-based and wide-field quantum noise spectroscopy to ground the article in current literature. Camera-based wide-field quantum noise spectroscopy (QNS) denotes a class of spatially resolved quantum measurement schemes in which a camera, SPAD array, CCD, EMCCD, sCMOS sensor, or event-based vision sensor acquires quantum-noise-sensitive observables over a large field-of-view in parallel, without the need of scanning operation. In the cited literature, the relevant observables include frame-by-frame quantum noise reduction of multi-spatial-mode twin beams, coincidence images from space-polarization hyper-entangled photon pairs, quadrature-noise variance maps of squeezed vacuum, full-field quantum illumination images through noise, and fluorescence-based readout of nitrogen-vacancy (NV) center spin protocols used to extract DC and AC magnetic field noise (Lawrie et al., 2013, Camphausen et al., 2021, Cuozzo et al., 2021, Gregory et al., 2019, Wang et al., 2023, Chen et al., 15 Sep 2025, Du et al., 2023).

1. Emergence of the field

A clear developmental trajectory runs from programmable single-pixel quantum imaging to genuinely wide-field, camera-native QNS. In 2013, a workbench for real-time quantum imaging measured frame-by-frame quantum noise reduction of multi-spatial-mode twin beams generated by four wave mixing in hot $^{85}\mathrm{Rb}$ vapor. Digital Micromirror Devices were used as spatial light modulators to pass arbitrary macropixels of quantum-correlated modes to a high quantum efficiency balanced detector, with the explicit aim of facilitating compressive quantum imaging with sensitivity below the photon shot noise limit. In 2016, the first sub-shot-noise wide field microscope based on spatially multi-mode non-classical photon number correlations in twin beams produced real time images of 8000 pixels at full resolution over a $(500\,\mu\mathrm{m})^2$ field-of-view, with noise reduced to the 80% of the shot noise level for each pixel. In 2021, a scan-free phase imager used a SPAD array camera to detect 2048 spatial modes in parallel, and in 2025 camera-based wide-field QNS with NV center spins in diamond explicitly implemented Rabi oscillation, Ramsey, Hahn echo and relaxometry experiments as the core of QNS (Lawrie et al., 2013, Samantaray et al., 2016, Camphausen et al., 2021, Chen et al., 15 Sep 2025).

This sequence is significant because it marks a shift in what is being parallelized. Early work parallelized spatial mode access through programmable masks and balanced detection; later work parallelized coincidence imaging and phase retrieval directly on a camera; NV-based work parallelized coherent spin protocols and noise-sensitive fluorescence readout across a field. The result is not a single technique but a convergent measurement paradigm in which quantum noise, correlations, or decoherence become image-forming quantities.

2. Platforms, sensors, and measured quantities

Camera-based wide-field QNS is not tied to a single quantum resource. The optical implementations use four-wave mixing twin beams, spatially multi-mode SPDC twin beams, space-polarization hyper-entanglement, squeezed vacuum with a local oscillator, and spatially correlated photon pairs in quantum illumination. Solid-state implementations use wide-field fluorescence readout from NV ensembles in diamond. The common architectural move is to replace serial interrogation by a pixelated or effectively pixelated measurement in which each camera pixel, macropixel, or coincidence map element becomes a spatial channel (Lawrie et al., 2013, Samantaray et al., 2016, Camphausen et al., 2021, Cuozzo et al., 2021, Gregory et al., 2019, Wang et al., 2023).

Platform	Detection architecture	Measured quantity
Four-wave-mixing twin beams in hot $^{85}\mathrm{Rb}$ vapor	DMD-programmed single-pixel balanced detection	Amplitude-difference noise, quantum noise reduction
Spatially multi-mode SPDC twin beams	CCD or EMCCD wide-field imaging	Photon-number correlations, sub-shot-noise absorption
Space-polarization hyper-entangled photon pairs	SPAD array camera	Coincidence images, phase-sensitive interference
Squeezed vacuum after an object	Camera in homodyne-like configuration	Quadrature-noise variance maps
SPDC quantum illumination	EMCCD full-field coincidence logic	Background rejection and contrast
NV ensemble diamond	SPAD array, industrial camera, sCMOS, event camera	ODMR, Rabi, Ramsey, Hahn echo, relaxometry

The measured observable depends on the platform. Twin-beam absorption microscopy subtracts symmetric pixels from probe and reference regions. The homodyne-like camera scheme subtracts corresponding pixels from two outputs of a beam splitter to mimic balanced homodyne difference current and access quadrature fluctuations. The SPAD-array phase imager extracts two-photon coincidences and phase-shifting digital holography observables. Quantum illumination retains only pixel pairs that satisfy a logical AND after the probe region is rotated by $\pi$. NV platforms record fluorescence changes during ODMR and coherent-control sequences, so QNS quantities are encoded in spatial maps of $T_1$, $T_2$, and $T_2^*$, or in fitted Rabi, Ramsey, Hahn-echo, and relaxometry contrasts.

3. Measurement models, estimators, and reconstruction

Across implementations, the central estimators are explicit. For the four-wave-mixing twin-beam source, squeezing was quantified as

$S = 10\log_{10} (1 - \eta - \eta / G),$

where $\eta$ is the net transmission and $G$ is the gain of the 4WM process. In the sub-shot-noise microscope, the noise reduction factor was

$(500\,\mu\mathrm{m})^2$0

In the homodyne-like camera protocol, the experimental variance for a bin of radius $(500\,\mu\mathrm{m})^2$1 centered at position $(500\,\mu\mathrm{m})^2$2 was

$(500\,\mu\mathrm{m})^2$3

and the reconstructed transmission map was

$(500\,\mu\mathrm{m})^2$4

In the SPAD-array phase imager, coincidence extraction used

$(500\,\mu\mathrm{m})^2$5

while the two-photon phase estimator was

$(500\,\mu\mathrm{m})^2$6

In NV-based wide-field QNS, the core observables are fitted contrast decays for Rabi, Ramsey, Hahn echo, and relaxometry, together with the filter-function expression

$(500\,\mu\mathrm{m})^2$7

These formulas show that camera-based QNS is not merely imaging with quantum light; it is estimator-driven measurement in which the camera outputs enter directly into variance, coincidence, or decoherence models (Lawrie et al., 2013, Samantaray et al., 2016, Cuozzo et al., 2021, Camphausen et al., 2021, Wang et al., 2023, Chen et al., 15 Sep 2025).

Reconstruction and optimal-measurement theory are equally central. The single-pixel twin-beam study cast compressive imaging as

$(500\,\mu\mathrm{m})^2$8

with the sub-Nyquist scaling $(500\,\mu\mathrm{m})^2$9, using an appropriate sparsifying transform $^{85}\mathrm{Rb}$0 and algorithms such as TV minimization. Tsang later formalized a correspondence between incoherent imaging and QNS through the random-displacement ensemble

$^{85}\mathrm{Rb}$1

showing that SPADE in imaging is analogous to spectral photon counting in optical phase noise spectroscopy and proposing unsqueezing before spectral photon counting for squeezed inputs. A plausible implication is that wide-field optical QNS and mode-resolved quantum imaging share not only instrumentation but also estimation-theoretic structure. On the computational side, kurtosis-difference weighted covariance introduced the excess kurtosis

$^{85}\mathrm{Rb}$2

the pairwise difference

$^{85}\mathrm{Rb}$3

and the weight

$^{85}\mathrm{Rb}$4

At 5000 frames, that method yielded a contrast-to-noise ratio exceeding 7, whereas standard covariance remained below 2, with a reported 40-fold reduction in acquisition time. A related proposal for quantum nonlinear spectroscopy uses the polarization of a coherent light beam as a pseudo-spin quantum sensor and states that a spatially resolved detector can provide a wide-field extension for higher-order time-ordered correlations, which suggests a route from camera-based noise mapping to camera-based higher-order QNS (Tsang, 2022, He et al., 30 Jun 2026, Cheung et al., 2023).

4. Wide-field operation, performance, and sensor bandwidth

Reported operating points span sub-shot-noise, single-photon, and coincidence-limited regimes. The four-wave-mixing twin-beam system reported noise reduction up to 4.5 dB below SNL for simple modes and more than 1 dB for patterned and compressively measured modes. The wide-field microscope achieved $^{85}\mathrm{Rb}$5 at full $^{85}\mathrm{Rb}$6 resolution and $^{85}\mathrm{Rb}$7 at $^{85}\mathrm{Rb}$8, corresponding to less than 30% of SNL after the resolution trade-off. The scan-free phase imager measured local uncertainty reductions from $^{85}\mathrm{Rb}$9 to $\pi$0 on a birefringent SLM pattern and from $\pi$1 to $\pi$2 on a protein microarray, together with an experimental ratio

$\pi$3

The homodyne-like camera scheme reconstructed an object with a total of 800 photons and less than one photon per frame on average, specifically $\pi$4 photons per pixel per frame. In full-field quantum illumination, the abstract reports image contrast improvement up to a factor of 5.5, while the detailed summary reports a background rejection ratio up to 5.8 and a quantum contrast advantage $\pi$5 up to 11, reflecting different performance metrics (Lawrie et al., 2013, Samantaray et al., 2016, Camphausen et al., 2021, Cuozzo et al., 2021, Gregory et al., 2019).

Camera technology determines the attainable temporal bandwidth and data volume. The SPAD-array wide-field quantum sensor used an MPD-SPC3 camera with $\pi$6 pixels, minimum integration time $\pi$7, dead time $\pi$8 ns, dark count rate $\pi$9 cps, maximum count rate $T_1$0 million counts per second per pixel, and frame rate up to $T_1$1 kHz. The neuromorphic widefield diamond sensor replaced frame readout by asynchronous spike encoding and experimentally demonstrated a 13x improvement in temporal resolution with comparable precision of detecting ODMR resonance frequencies; the detailed comparison reports 35 MB per ODMR scan for EMCCD versus 363 KB for the event camera, and resonance-frequency precision of $T_1$2 MHz versus $T_1$3 MHz (Wang et al., 2023, Du et al., 2023).

5. NV-center wide-field QNS and materials-oriented implementations

NV-center platforms convert camera-based wide-field QNS from an optical-imaging analogy into a direct spectroscopy of local environments. A SPAD-array implementation experimentally demonstrated sensing DC and AC magnetic fields, temperature, strain, local spin density, and charge dynamics using an NV ensemble diamond sample. The 2025 camera-based QNS work then addressed the specific conflict between the optimal optical spin readout time, normally below one microsecond, and the minimal camera exposure time, normally tens of microsecond. Using a home-built setup, it demonstrated fully coherent control via Rabi oscillation, Ramsey, Hahn echo and relaxometry experiments, achieved an unexpectedly high contrast of 12% with an optical spin readout time of 500 microseconds, and reported ODMR contrast up to 13.6% with 500 microseconds exposure. The proposed mechanism is weak laser illumination, for which spin-lattice relaxation and weak optical pumping cause a rather slow reduction of contrast with increasing integration time; the abstract states that this is instructive for constructing wide-field QNS with a sCMOS or EMCCD camera and for studies of magnetic material or superconducting material (Wang et al., 2023, Chen et al., 15 Sep 2025).

The solid-state camera literature also emphasizes that QNS is constrained by readout architecture as much as by quantum control. The event-based widefield ODMR system showed that fluorescence changes near resonance can be encoded as timestamped spikes rather than full frames, and demonstrated monitoring of dynamically modulated laser heating of gold nanoparticles coated on a diamond surface. A plausible implication is that camera-based wide-field QNS can be divided into frame-based, photon-counting, and event-based regimes, with the choice of regime set by whether the dominant bottleneck is readout noise, coincidence extraction, or data transfer and latency (Du et al., 2023).

6. Limitations, misconceptions, and extensions

Several persistent misconceptions are directly addressed in the literature. One is that wide-field quantum enhancement necessarily requires scanning: the SPAD-array phase imager acquired all image pixels simultaneously with no moving parts or scanning required, while the 2016 wide-field microscope was already a real-time full-field system. Another is that camera exposure times fundamentally preclude NV-spin QNS; the weak-illumination readout regime demonstrates otherwise. A third is that direct intensity imaging and homodyne-type measurements are generically optimal; in Tsang’s analysis, SPADE and unsqueezed spectral photon counting can achieve the respective quantum limits and are far superior to direct displacement measurements such as direct imaging or homodyne detection in the relevant regimes (Camphausen et al., 2021, Samantaray et al., 2016, Chen et al., 15 Sep 2025, Tsang, 2022).

The principal limitations are equally explicit. Losses in the beam path degrade quantum correlations, so SLM throughput matters in twin-beam schemes. The homodyne-like camera method was designed precisely because camera dark noise is detrimental in the weak-illumination regime. In spectral-photon-counting analogs, the unsqueezing process must be quantum-limited, and practical optical squeezers may introduce additional noise while requiring two squeezers. Photon-sparse SPDC imaging can demand very large frame counts: the full-field quantum-illumination experiment accumulated several million frames, standard covariance methods in SPDC imaging typically require tens of thousands of frames, and the kurtosis-weighted method was introduced to reduce that burden. A related direction proposes using weak Faraday-rotation measurements of a coherent light beam, with a spatially resolved detector for wide-field extension, to extract arbitrary types and orders of time-ordered correlations; this suggests that camera-based wide-field QNS may expand from second-order noise characterization toward systematic quantum nonlinear spectroscopy (Lawrie et al., 2013, Cuozzo et al., 2021, Gregory et al., 2019, He et al., 30 Jun 2026, Cheung et al., 2023).