Author Affiliations
Key Laboratory of Optoelectronic Imaging Technology and Systems, Ministry of Education, School of Optics and Photonics, Beijing Institute of Technology, Beijing 100081, Chinashow less
Fig. 1. Schematic of the speckle imaging system
[44] Fig. 2. Recursion-driven bispectral imaging (ReDBI) framework
[45] Fig. 3. Reconstruction of stationary objects hidden behind a dynamic scattering medium via ReDBI
[45]. (a) Raw 512×512 pixel-sized speckle images of scattered objects' light; (b)-(f) reconstruction results when using different number of 512×512 pixel-sized speckle images; (g) original object patterns
Fig. 4. Reconstruction of moving objects hidden behind a dynamic scattering medium via ReDBI
[45]. (a) Schematic of the object moving to seven different positions on the object plane when capturing speckle images; (b)-(f) reconstruction results when using different number of 512×512 pixel-sized speckle images at each position
Fig. 5. Deep learning dynamic target imaging and tracking through scattering media driven by speckle difference
[44]. (a) Schematic of the proposed learning method; (b) reconstruction and tracking results for moving objects
Fig. 6. Reconstruction results of moving human faces and objective evaluation results on FEI Face dataset
[44]. (a) Reconstruction results; (b) corresponding MAE, SSIM, and PSNR
Fig. 7. Comparison of reconstruction results of PnPGAP-FPR and FPR
[46]. (a)(b) Speckle patterns captured under the darkroom scene, the sub-window shows the ground truth; (i) visualization results and corresponding mean values under different noise levels; (ii)(iv) results restored via PnPGAP-FPR; (iii)(v) results restored via FPR
Fig. 8. [in Chinese]
Fig. 9. Actual FPA CI system in MWIR and super-resolution reconstruction images of a temperature-controlled electric iron with different compression ratios
[49] Fig. 10. Specific details of the previously mentioned three methods. (a) High-resolution MWIR images are reconstructed with different compression ratios
[51]; (b) reconstruction of electric soldering iron images by using different methods
[52]; (c) proposed SLC method
[53] Fig. 11. High-resolution fast mid-wave infrared compressive imaging
[55]. (i) Template calibration effect, wherein (a) is original mask, (b) is measured mask, (c) is uniformly calibrated mask, (d) is deconvoluted mask, and (e) is nonuniformly calibrated mask; (ii) sliding window measurement collection processing; (iii) wherein (a) is low-resolution image, (b) is uncalibrated reconstruction (complete video in Visualization 1), (c) is deconvoluted calibrated reconstruction, (d) is nonuniform calibrated reconstruction (complete video in Visualization 2), with the sampling rate of 12.5%
Fig. 12. Architecture of Meta-TR
[56] Fig. 13. SCI system and Joinput-CiNet framework, simulated four-bar targets reconstructed by using Joinput-CiNet and ReconNet
[57] Fig. 14. 3D-TCI-CNN
[61]. (a) Structure of 3DTCI (3D-TCI-CNN); (b) structure of 3DTCI-R4 (3D-TCI-CNN with four 3D-TCI-R4 units); (c) results of reconstructing moving / rotating targets by using the 3DTCI-R4
Fig. 15. Broadband dual-band TCI experimental device
[63] Fig. 16. GapUNet
[66]. (a) Network structure; (b) simulation results; (c) reconstruction results in optical experiments
Fig. 17. Reflective optical off-axis CASSI system. (a) Schematic of the system; (b) experimental system (revised from Fig. 2 and Fig. 3 in Ref.[
67])
Fig. 18. DD-CASSI experimental system (revised from Fig. 5 in Ref.[
68])
Fig. 19. Hyperspectral image classification results of different methods (revised from Fig. 11 in Ref. [
69])
Fig. 20. Hexagonal blue noise complementary coded aperture (revised from Fig. 1 and Fig. 6 in Ref.[
70]). (a) Sketch of CASSI system with hexagonal blue noise complementary coded aperture; (b) hexagonal blue noise coded aperture with transmittance of 10%; (c) hexagonal blue noise coded aperture with transmittance of 16.67%
Fig. 21. Adaptive coded aperture according to space scene (revised from Fig. 1 and Fig. 5 in Ref.[
71]). (a) Space scene; (b) adaptive coded aperture; (c) comparison of reconstruction results between adaptive coding and other coding methods
Fig. 22. Fama-SDIP reconstruction framework and experimental results of CASSI system (revised from Fig. 1, Fig. 3, Fig. 5, Fig. 6, and Fig. 7 in Ref.[
73]). (a) Imaging process of CASSI system; (b) diagram of deep image prior network structure; (c) experimental system; (d) comparison of simulation results of different methods; (e) comparison of experimental results of different methods