Fig. 1. Schematic diagram of light field imaging with achromatic metalens array and rendered images^[32]

Fig. 2. Schematic diagram of VMMA^[36]. Working principle of VMMA when linearly polarized light on the (a) x-axis (TE) and (b) y-axis (TM) is incident; (c) configuration of a single metasurface constituting the VMMA, including unit cells of different types of nanopillar; simulated phase distribution of 2×2 VMMA with (d) TE and (e) TM polarized incident light; simulated far-field intensity distribution of (f) TE and (g) TM incident light

Fig. 3. Three metasurfaces share apertures to realize metasurface light fields for single particle tracking^[38]

Fig. 4. PMDP metalens array^[37]. (a) Phase distribution of PMDP metalens array in the x-y plane, while the blue and red squares denote the phase distribution method for LCP and RCP respectively; (b) two raw images captured by LCP and RCP illumination respectively; (c) stitched result from Fig. (b) through a certain image processing

Fig. 5. Independent dual-phase design for co- and cross-circularly polarized light^[29]

Fig. 6. Metalens array planar wide-angle camera^[30]. (a) Each part of FoV is imaged separately by each metalens; (b) schematic illustration of focusing oblique incident light by a metalens

Fig. 7. Light field imaging system enabled by the spin-multiplexed metalens array^[31]. (a) Conceptual sketch of the proposed light field imaging camera; (b) bioinspired photonic spin-multiplexed metalens array with its nanopillar unit cell; (c) schematic diagram of the working principle of the system

Fig. 8. Bionic metalens depth sensing system utilizing the lateral dual focus^[39]

Fig. 9. Dual-aperture metasurface utilizing depth-invariant and depth-sensitive PSF simultaneously for depth imaging^[34]

Fig. 10. Disparity (depth) computation with the neural network^[54]. (a) Architecture overview of proposed neural network H-Net with H-Module; (b) detailed pipelines of the cross-pixel interaction; (c) detailed pipelines of cross-view interactions

Fig. 11. Schematic of the achromatic metalens array depth-sensing system^[35]

Fig. 12. Relationship between simulated deflection efficiency and deflection angle of a large-angle, multifunctional metasurface grating designed using topology optimization in (a) theory and (b) experiment^[71]

Fig. 13. Metasurface designed using topology optimization with a target refraction angle of 20° and a target phase of π/2^[72]

Fig. 14. Genetic algorithm based design of high focusing efficiency metalens^[73]. (a) Electric energy density of the ideal lens; (b) genetic algorithm optimization flowchart; (c) layout of the non-periodic metasurface lens designed

Fig. 15. 3D plasma metasurface design based on ant colony algorithm^[74]

Fig. 16. (a) Actual phase distribution, (b) target phase distribution, and (c) corresponding top-view layout of the metasurface designed using the particle swarm optimization algorithm^[75]

Fig. 17. DNN-based optimization of metalens parameters^[81]

Fig. 18. Metasurface design based on forward and inverse neural networks^[82]

Fig. 19. Multilayer metasurface design based on GAN^[86]

Fig. 20. Architecture of the metasurface design network based on WGAN^[88]. (a) Schematic of the network structure; (b) flowchart of the structural process; (c) example of the design process; (d) schematic illustration of the final metasurface structure designed

Fig. 21. End-to-end imaging pipeline composed of the metasurface image formation model and the feature-based deconvolution algorithm^[91]

Fig. 22. Learned metalens array for capturing wide FoV light field^[92]

Fig. 23. Architecture of proposed end-to-end designed MLF

Fig. 24. Calculation of depth of field range of ideal lens imaging. (a) Schematic diagram of object point imaging closer than the focal point; (b) schematic diagram of object point imaging further than the focal point

Fig. 25. Architecture of multi-scale image fusing network

Fig. 26. Qualitative assessment of experimental imaging results

Fig. 27. Zoom-in comparation between ground truth images and E2EMLF results