Computational Ghost Imaging Techniques

Computational ghost imaging encompasses a suite of indirect imaging methods that reconstruct spatial or temporal scenes from correlations between known illumination patterns and intensity measurements recorded by a single-element detector. Rather than relying on a pixelated sensor array, the technique projects a sequence of spatial or temporal modulation patterns—often generated by a spatial light modulator—onto an object and measures the total transmitted or reflected intensity, known as a bucket signal. Numerical algorithms then invert these correlations to recover the image. Advances in compressed sensing have dramatically reduced the number of measurements required by exploiting sparsity in natural scenes, while machine-learning approaches have further enhanced reconstruction quality under photon-starved or sub-Nyquist sampling conditions. Extensions into three dimensions, hyperspectral regimes and the temporal domain have broadened the technique’s scope, enabling depth mapping, mid-infrared and terahertz imaging, and ultrafast dynamics studies. The inherent flexibility, low cost and resilience to environmental noise make computational ghost imaging attractive for remote sensing, biomedical diagnostics and industrial inspection, especially where conventional detectors are unavailable or impractical.

Research from Nature Portfolio

Recent studies have demonstrated single-photon-level ghost imaging in the mid-infrared by imprinting time-varying pump patterns onto an infrared object via nonlinear sum-frequency generation, translating the spectral content into the visible for high-sensitivity detection. Compressed sensing and deep-learning algorithms then accurately reconstruct images from sub-Nyquist and photon-sparse measurements, offering a new route to sensitive long-wavelength and terahertz imaging. Another seminal work employed a deep neural network trained on pairs of ghost-imaged reconstructions and ground-truth scenes to learn the underlying sensing model. This framework markedly improves image fidelity at extremely low sampling rates, outperforming traditional compressed-sensing methods in both simulation and optical experiments.

Computational Ghost Imaging Techniques publication trend

The graph below shows the total number of articles in computational ghost imaging techniques across all publications each year (not limited to Nature Index journals).

Technical terms

Ghost imaging: Reconstruction of an image by correlating known illumination patterns with integrated intensity measurements from a single-element detector.

Single-pixel camera: Imaging system that projects sequential spatial patterns onto a scene and records the total transmitted or reflected light with a single detector pixel.

Compressed sensing: Signal-processing technique that recovers high-fidelity images from undersampled data by exploiting sparsity in a suitable transform domain.

Bucket detector: Single-element photodetector that measures the aggregate light intensity after interaction with the object.

Spatial light modulator: Programmable device used to impose structured illumination patterns onto a light beam.

References

1. Cascaded compressed-sensing single-pixel camera for high-dimensional optical imaging. PhotoniX (2024).
2. Mid-infrared computational temporal ghost imaging. Light: Science & Applications (2024).
3. Mid-infrared single-pixel imaging at the single-photon level. Nature Communications (2023).
4. Single-pixel imaging 12 years on: a review.. Optics Express (2020).
5. Deep-learning-based ghost imaging. Scientific Reports (2017).