Author Affiliations
1Department of Electronic Engineering, Tsinghua University, Beijing 100084, China2Beijing National Research Center for Information Science and Technology (BNRist), Beijing 100084, China3Beijing Academy of Quantum Information Science, Beijing 100084, Chinashow less
Fig. 1. Classification of spectral imaging by acquisition methods
[11]. (a) Scanning spectral imaging; (b) snapshot spectral imaging
Fig. 2. Principle of metasurface based spectral imaging. (a) Structure diagram of metasurface spectral imaging chip, including two parts of metasurface layer and CMOS image sensor; (b) principle of spectral reconstruction for a single metasurface microspectrometer; (c) principle of space division multiplexing of metasurface
Fig. 3. Design objective of metasurface units. (a) Projection of spectral lines at onto transmission spectra of metasurface units; (b) corresponding measurement vectors for spectral lines at are two columns of measurement matrix without considering measurement noise
Fig. 4. World's first real-time ultraspectral imaging chip and its performance
[17]. (a) Schematic diagram of chip structure; (b) reconstruction results of narrowband spectra using microspectrometer of chip; (c) center-wavelength error and linewidth error for results in (b); (d) reconstruction results of double-peak signal
Fig. 5. Ultraspectral imaging chip based on metasurfaces with freeform shaped meta-atoms
[19]. (a) Schematic diagram of structure; (b) reconstruction results of double-peak signal
Fig. 6. Spectral imaging results for a standard colorboard and a plate of fruit using ultraspectral imaging chip based on metasurfaces with freeform shaped meta-atoms
[19]. (a) Object picture of ultraspectral camera; (b) red, green, and blue (RGB) pseudo-color image of standard colorboard captured by a commercial spectral camera; (c) RGB pseudo-color image of standard colorboard reconstructed by ultraspectral camera; (d) reconstructed spectra (red lines) for 24 types of colors with spectra captured by commercial spectral camera as a reference (blue lines), in which fidelities for recovered spectra are marked in top right-hand corner; (e) spectral imaging results for a plate of fruit and spectral reconstruction results for sampling points
Fig. 7. Basic architecture of ADMM-net and reconstruction results of a standard colorboard using ADMM-net
[26]. (a) Network structure of ADMM-net; (b) spectral image reconstruction results of a standard colorboard using ADMM-net
Fig. 8. Real-time brain spectral imaging results for a rat using world's first real-time superspectral imaging chip
[17]. (a) Spectral imaging results; (b) object picture of spectral camera; (c) spectral signals of different regions; (d) changes of spectral signals of HbO and HbR over time in vascular areas; (e) changes of spectral signals of HbO and HbR over time in non-vascular areas
Fig. 9. Snapshot spectral measurement results of living faces and common camouflage materials
[10]. (a) Live face; (b) paper mask; (c) silicone mask; (d) raw silicone material
Fig. 10. Real-time spectral imaging results of an outdoor driving scene
[26] Method | Line scanning | ADMM-net | GAP-TV | λ-net | CVX |
---|
Data cube size | 256×256×26 | 256×256×601 | 256×256×26 | Running time /s | ~60 | 1.72 @CPU 0.018 @GPU | 110 @CPU | 2.44 @CPU 0.095 @GPU | 7767 @CPU | 4854 @CPU |
|
Table 1. Running time of different spectral imaging methods