• Infrared and Laser Engineering
  • Vol. 53, Issue 9, 20240378 (2024)
Tiancheng WANG1, Wangtao YU1,2, Weiyun CHEN1, and Zhongyi GUO1
Author Affiliations
  • 1School of Computer and Information, Hefei University of Technology, Hefei 230009, China
  • 2Sun Create Electronics Co. LTD., Hefei 230094, China
  • show less
    DOI: 10.3788/IRLA20240378 Cite this Article
    Tiancheng WANG, Wangtao YU, Weiyun CHEN, Zhongyi GUO. Research advances on Fourier single-pixel imaging technology (invited)[J]. Infrared and Laser Engineering, 2024, 53(9): 20240378 Copy Citation Text show less
    Schematic diagram of the SPI. (a) The active SPI; (b) The passive SPI
    Fig. 1. Schematic diagram of the SPI. (a) The active SPI; (b) The passive SPI
    Schematic diagram of the synthesis and decomposition of 2D image based on Fourier forward and inverse transforms [37]
    Fig. 2. Schematic diagram of the synthesis and decomposition of 2D image based on Fourier forward and inverse transforms [37]
    Schematic diagram of the FSPI experimental system [38−39]
    Fig. 3. Schematic diagram of the FSPI experimental system [3839]
    (a) The temporal dithering strategy; (b) The spatial dithering strategy; (c) The reconstructed Fourier spectra and images using different binarization methods[40]
    Fig. 4. (a) The temporal dithering strategy; (b) The spatial dithering strategy; (c) The reconstructed Fourier spectra and images using different binarization methods[40]
    (a) The temporal dithering strategy; (b) The signal dithering strategy; (c) The reconstructed images under different spectral coverage[41]
    Fig. 5. (a) The temporal dithering strategy; (b) The signal dithering strategy; (c) The reconstructed images under different spectral coverage[41]
    (a) Error diffusion kernels of Floyd-Steinberg and Zhang-Qi; (b) Eight different image dithering scanning strategies; (c) Different scanning methods (top row) and corresponding partial amplification (bottom row) based on Zhang-Qi dithering strategies; (d) The experimental results of 3D target scene[42]
    Fig. 6. (a) Error diffusion kernels of Floyd-Steinberg and Zhang-Qi; (b) Eight different image dithering scanning strategies; (c) Different scanning methods (top row) and corresponding partial amplification (bottom row) based on Zhang-Qi dithering strategies; (d) The experimental results of 3D target scene[42]
    (a) Schematic diagram based on Gaussian random sampling [44]; (b) The reconstruction results at different DMD refresh rates [44]; (c) Schematic diagram based on sparse sampling [45]; (d) The reconstruction results under different sampling methods [45]
    Fig. 7. (a) Schematic diagram based on Gaussian random sampling [44]; (b) The reconstruction results at different DMD refresh rates [44]; (c) Schematic diagram based on sparse sampling [45]; (d) The reconstruction results under different sampling methods [45]
    (a) Flowchart of adaptive FSPI based on Fourier domain radial correlation [49]; (b) Schematic diagram of the adaptive FSPI based on DQN [50]; (c) Adaptive FSPI scheme based on the AuSamNet [51]
    Fig. 8. (a) Flowchart of adaptive FSPI based on Fourier domain radial correlation [49]; (b) Schematic diagram of the adaptive FSPI based on DQN [50]; (c) Adaptive FSPI scheme based on the AuSamNet [51]
    (a) Schematic diagram of multi-block FSPI via frequency division multiplexed modulation [52]; (b) The reconstruction results of the target at a sampling rate of 1% under different signal-to-noise ratios [52]; (c) Schematic diagram of SPI for high performance based on optimized sinusoidal patterns [54]; (d) The reconstruction results of the target at a sampling rate of 20% under different signal-to-noise ratios [54]
    Fig. 9. (a) Schematic diagram of multi-block FSPI via frequency division multiplexed modulation [52]; (b) The reconstruction results of the target at a sampling rate of 1% under different signal-to-noise ratios [52]; (c) Schematic diagram of SPI for high performance based on optimized sinusoidal patterns [54]; (d) The reconstruction results of the target at a sampling rate of 20% under different signal-to-noise ratios [54]
    (a) Representation of nonlocal 3D sparsity in the transform domain; (b) The reconstruction results of different algorithms; (c) The experimental reconstruction results under different sampling rates[55]
    Fig. 10. (a) Representation of nonlocal 3D sparsity in the transform domain; (b) The reconstruction results of different algorithms; (c) The experimental reconstruction results under different sampling rates[55]
    (a) The generative adversarial network architecture [59]; (b) Comparison of reconstruction results under different methods [59]; (c) Schematic diagram of high-resolution iterative reconstruction based on diffusion model [60]; (d) Comparison of reconstruction methods under different sampling rates [60]
    Fig. 11. (a) The generative adversarial network architecture [59]; (b) Comparison of reconstruction results under different methods [59]; (c) Schematic diagram of high-resolution iterative reconstruction based on diffusion model [60]; (d) Comparison of reconstruction methods under different sampling rates [60]
    (a) Schematic diagram of dual-contour data processing based on the FSPI [67]; (b) Edge detection method based on the FSPI [76]
    Fig. 12. (a) Schematic diagram of dual-contour data processing based on the FSPI [67]; (b) Edge detection method based on the FSPI [76]
    [77] (a) Schematic diagram of polarized computational ghost imaging in scattering system with half-cyclic sinusoidal patterns; (b) The reconstruction results of different methods in the same scattering system
    Fig. 13. [77] (a) Schematic diagram of polarized computational ghost imaging in scattering system with half-cyclic sinusoidal patterns; (b) The reconstruction results of different methods in the same scattering system
    Feature/StrategyTemporal ditheringSpatial ditheringSignal dithering
    Temporal resolutionLowHighMiddle
    Spatial resolutionHighLowHigh
    Table 1. Comparison of three different dithering strategies
    Feature/StrategyTraditional samplingSparse samplingAdaptive sampling
    Sampling modeFixedVariable densityAdaptive
    AdvantageSimplenessHigh qualityHigh quality, generalization
    DisadvantagedPoor flexibilityHigh complexityConsume resource
    ScenariosRegular structureGeneralReal-time
    Table 2. Comparison of different sampling-path strategies
    Feature/StrategySecond-order correlationCompressed sensingInverse Fourier transformDeep learning
    MethodStatistical averagingSparsity optimizationFrequency domain to spatial domainFeature extraction optimization
    EfficiencyMiddleLowMiddleHigh
    QualityMiddleHighHighHigh
    Table 3. Comparison of different reconstruction algorithms
    Tiancheng WANG, Wangtao YU, Weiyun CHEN, Zhongyi GUO. Research advances on Fourier single-pixel imaging technology (invited)[J]. Infrared and Laser Engineering, 2024, 53(9): 20240378
    Download Citation