Quantum Visual Field (QVF)

• Quantum Visual Field (QVF) is a framework that defines the spatial sensitivity of visual systems via quantum-limited detection and implicit neural representations.
• It integrates correlation-based imaging methods with parameterized quantum circuits to reconstruct high-resolution maps in both biological and artificial vision.
• QVF research highlights practical advances in biomedical imaging and quantum machine learning while addressing challenges like device scaling and noise management.

A Quantum Visual Field (QVF) is a foundational construct at the intersection of quantum optics, neural field representations, and sensory biophysics. QVF refers both to (1) the spatial map of quantum-limited sensitivity in the human or biological visual system, determined by Poissonian detection theory and correlation-based quantum imaging, and (2) a class of quantum implicit neural representations (QINRs) that encode multidimensional classical fields—such as images or 3D scenes—into quantum state vectors using amplitude encoding and parameterized quantum circuits. The QVF unites diverse methodologies: quantum measurement theory in anatomy and psychophysics, second-order correlation (ghost imaging), and continuous coordinate-based field learning on quantum devices. This article surveys the core theoretical structure of the QVF, its mathematical models, experimental and algorithmic realizations, and emerging practical applications in biomedical imaging and quantum machine learning.

1. Quantum-Limited Detection and Visual Perception

The biophysical foundation of QVF in vision science is the quantum-limited nature of rod photoreceptors, which operate as Poissonian photon counters in the photon-starved regime. For a rod receiving a mean incident photon number $\mu$ in an integration window $\Delta t$, the photon arrival statistics are governed by $P(n|\mu) = e^{-\mu}\frac{\mu^n}{n!}$. Detection proceeds with quantum efficiency $\eta$, so the probability of at least one absorbed photon (the "click" probability) is $P_\mathrm{click} = 1 - e^{-\eta\mu}$. The threshold for perceptual detection, $\mu_\mathrm{th}$, is defined by solving $1 - e^{-\eta\mu_\mathrm{th}} = p_0$, exemplarily for $p_0 = 0.5$, yielding $\mu_\mathrm{th} = -\frac{1}{\eta}\ln(1-p_0)$. Rods also experience dark isomerizations at rate $d$ per $\Delta t$0, so total noise is $\Delta t$1. In the low-flux limit, the system is governed by Poisson noise, and the signal-to-noise ratio (SNR) for $\Delta t$2 rods (or pulses) is $\Delta t$3.

Spatial resolution in the quantum-limited regime is set by receptor spacing $\Delta t$4 and the eye’s focal length $\Delta t$5, with the smallest resolvable angle $\Delta t$6. Foveal spacing $\Delta t$7m yields $\Delta t$8 arcmin. Temporal summation windows, $\Delta t$9 ms, cause closely spaced pulses to be perceptually integrated.

These detection-theoretic results underpin a physical, noise-limited conception of the visual field: the spatial domain over which visual signals can be detected with a prescribed probability, as a function of incident photon budget, integration window, and spatial pattern (Kulmaganbetov et al., 17 Jun 2026).

2. Correlation-Based Imaging and Quantum Retinal Mapping

QVF is quantitatively and operationally realized via correlation-based imaging, most prominently ghost imaging using quantum or classical light. Stimulating the retina with a family of known spatial patterns $P(n|\mu) = e^{-\mu}\frac{\mu^n}{n!}$0 and measuring integrated detection responses $P(n|\mu) = e^{-\mu}\frac{\mu^n}{n!}$1 across the region, the retinal transmission or sensitivity at point $P(n|\mu) = e^{-\mu}\frac{\mu^n}{n!}$2 is reconstructed as

$P(n|\mu) = e^{-\mu}\frac{\mu^n}{n!}$3

In the quantum-optical (second-order correlation) language, this is expressed via $P(n|\mu) = e^{-\mu}\frac{\mu^n}{n!}$4, with normalized ghost images $P(n|\mu) = e^{-\mu}\frac{\mu^n}{n!}$5. By scanning across the stimulus set and recording detection events at each retinal location, one reconstructs a high-resolution local map of quantum-limited detection efficiency $P(n|\mu) = e^{-\mu}\frac{\mu^n}{n!}$6, providing the operational definition of the spatial extent and granularity of the QVF. Experiments employing structured-light fields (e.g., vector-vortex beams with orbital angular momentum) further probe angular QVF boundaries, mapping functional sensitivity profiles under both standard and nontrivial quantum optical stimuli (Kulmaganbetov et al., 17 Jun 2026).

3. QVF in Quantum Machine Learning: Amplitude Encoding and Field Learning

The QVF paradigm in quantum machine learning formulates the field as a coordinate-based quantum implicit neural representation. A QVF learns a waveform or image $P(n|\mu) = e^{-\mu}\frac{\mu^n}{n!}$7 on input $P(n|\mu) = e^{-\mu}\frac{\mu^n}{n!}$8 by mapping it through a parameterized quantum circuit, yielding a quantum state $P(n|\mu) = e^{-\mu}\frac{\mu^n}{n!}$9. The decoding proceeds directly via projective local Pauli-$\eta$0 measurements to estimate the field value $\eta$1, trained to match $\eta$2 under a loss,

$\eta$3

The QVF architecture decomposes as follows (Wang et al., 14 Aug 2025, Bosco et al., 8 Jan 2026):

• A coordinate encoding network $\eta$4 (MLP) produces an energy spectrum $\eta$5 conditional on spatial position and optional latent codes;
• Boltzmann-regulated amplitude encoding transforms $\eta$6 into normalized amplitudes $\eta$7 for the $\eta$8 computational basis states;
• A fully entangled variational circuit $\eta$9, entirely in the real Hilbert subspace, transforms the amplitudes;
• Projective $P_\mathrm{click} = 1 - e^{-\eta\mu}$0-basis measurements yield the predicted field at the desired location;
• The parameter-shift rule enables end-to-end differentiable training through both classical and quantum parameters.

The quantum-only readout (without heavy classical postprocessing) enables stable, resolution-agnostic field approximation. QVF admits extension to conditional and multimodal data via latent variables and supports completion and interpolation tasks in 2D and 3D fields (Wang et al., 14 Aug 2025).

4. Experimental Platforms and Empirical Metrics

QVF approaches have been implemented at the interface of physics and machine learning, as well as in laboratory experimental biophysics. In quantum neural applications (Wang et al., 14 Aug 2025, Bosco et al., 8 Jan 2026), QVF is empirically validated by fitting to high-resolution 2D images (CIFAR-10, JWST datasets) and 3D signed-distance fields (ShapeNet). Metrics include mean-squared error (MSE), peak signal-to-noise ratio (PSNR), and mean absolute error (MAE) for field completion tasks. QVF demonstrates enhanced high-frequency detail learning and lower reconstruction error compared to classical MLP and neural radiance fields (NeRF) baselines. For example, single-image QVF (with 5 qubits, depth 5) achieves MSE $P_\mathrm{click} = 1 - e^{-\eta\mu}$1 and PSNR $P_\mathrm{click} = 1 - e^{-\eta\mu}$2 dB, outperforming QIREN and classical MLPs.

QVF-augmented neural radiance field models ("QNeRF") achieve parameter savings of over 60% over classical NeRFs at comparable or better PSNR on synthetic and real novel-view synthesis datasets. Full QNeRF attains PSNR $P_\mathrm{click} = 1 - e^{-\eta\mu}$3 dB using only 222k parameters (classical NeRF: 590k), with further reduction and hardware compatibility in dual-branch QNeRF architectures (Bosco et al., 8 Jan 2026).

In biophysical settings, QVF maps have been empirically measured using programmable, Poisson-distributed, spatially selective photon sources scanned across the retina or pupil, with sensitivity mapped pixel-by-pixel via psychometric or involuntary pupillometric responses and thresholds (Margaritakis et al., 2019, Kulmaganbetov et al., 17 Jun 2026).

5. Visualization, Intuitive Interpretation, and Quantum Field Theory

A pedagogically motivated construction of QVF in the context of quantum field theory visualizes the many-body wave function $P_\mathrm{click} = 1 - e^{-\eta\mu}$4 for discretized bosonic fields as an ensemble of randomly sampled classical oscillator configurations $P_\mathrm{click} = 1 - e^{-\eta\mu}$5, each weighted and color-mapped by the value $P_\mathrm{click} = 1 - e^{-\eta\mu}$6 in normal-mode coordinates. This methodology, involving parallel-coordinate plots and grayscale (or color) amplitude mappings, exposes field excitations, spatial correlations, vacuum fluctuations, and nodal hypersurfaces, contributing to the "field intuition" bridging first- and second-quantized descriptions (Linde, 2019).

The following table summarizes canonical QVF realizations across domains:

Domain	QVF Implementation	Key Metric / Observable
Human vision psychophysics	Poisson-threshold map, ghost imaging	$P_\mathrm{click} = 1 - e^{-\eta\mu}$7, SNR
Quantum machine learning (QINR)	Amplitude encoding, entangled PQC	MSE/PSNR/MAE on fields
Wide-field quantum imaging	SPAD N00N-state phase mapping	Sensitivity enhancement ($P_\mathrm{click} = 1 - e^{-\eta\mu}$8)
Bosonic QFT pedagogy	Ensembles of field configurations	Visualization of $P_\mathrm{click} = 1 - e^{-\eta\mu}$9

6. Limitations, Technical Challenges, and Future Directions

Current QVF approaches in both vision science and quantum neural fields are constrained by device scaling and noise. In QINRs, classical simulation cost grows as $\mu_\mathrm{th}$0, limiting quantum circuit width to $\mu_\mathrm{th}$1 qubits; all experiments to date use simulators rather than physical quantum hardware (Wang et al., 14 Aug 2025, Bosco et al., 8 Jan 2026). In imaging, loss-sensitive quantum states (e.g., N00N) degrade, but advances in single-photon detection efficiency and high-frame-rate SPAD arrays mitigate this. In physiological mapping, spatial resolution is limited by detector density and eye movement control, though feedback-stabilized, IR-registered, and deep-learning–guided systems are enhancing spatial precision (Margaritakis et al., 2019).

Anticipated future directions include:

• Efficient, scalable preparation of learnable quantum Gibbs states (tensor-network methods);
• Hybrid quantum-classical pipelines for neural field rendering and 2D→3D completion;
• Adaptation to multi-scale and dynamic scene representation within QVF-augmented NeRF frameworks;
• Hardware-adapted ansätze and error mitigation for noisy intermediate-scale quantum (NISQ) devices;
• Analytical advances in quantum detection theory for complex, structured-light and low-photon-budget vision experiments (Wang et al., 14 Aug 2025, Bosco et al., 8 Jan 2026, Kulmaganbetov et al., 17 Jun 2026).

A plausible implication is that QVF formalism provides a unifying operational, mathematical, and algorithmic toolset for exploring both biological and artificial quantum-enhanced field representations, with applications spanning biomedicine, quantum imaging, and machine perception.

7. Conceptual Synthesis and Significance

The QVF constitutes the high-resolution, photon-budget–limited sensitivity map—implemented or measured via quantum detection theory, correlation imaging metrics, and continuous quantum learning architectures—of a visual (or data) system. In biophysics, QVF operationalizes the quantum limits of perception and underlies advanced diagnostics in ophthalmology. In quantum neural representation, it enables continuous, resolution-agnostic field learning with demonstrable parameter efficiency and high fidelity. QVF methodology spans and connects quantum biology, imaging, and machine learning, providing a rigorous framework for interrogating physical and algorithmic limits of information extraction from visual fields (Kulmaganbetov et al., 17 Jun 2026, Wang et al., 14 Aug 2025, Bosco et al., 8 Jan 2026, Margaritakis et al., 2019, Camphausen et al., 2021, Linde, 2019).