The thesis explores three challenges in 3D imaging with different applications: 3D stereo imaging with large depth-of-field, 3D sensing with a compact device, and 3D microscopy of thick scattering samples with fast scanning speed.

The first part of this thesis focuses on a stereo imaging system that can get large depth-of-field and high-quality 3D reconstruction in light-limited environments. To overcome the fundamental trade-off between imaging volume and signal-to-noise ratio (SNR) that appears in conventional stereo, a novel end-to-end learning-based technique is proposed by introducing a phase mask at the aperture plane of the cameras in a stereo imaging system. The phase mask creates a depth-dependent yet numerically invertible point spread function, allowing us to recover sharp image texture and stereo correspondence over a significantly extended depth of field (EDOF) than conventional stereo.

The second part of the thesis exploits the strongly dispersive property of metasurfaces to propose a compact, single-shot, and passive 3D imaging camera. The proposed device consists of a metalens engineered to focus different wavelengths at different depths and two deep networks to recover depth and RGB texture information from chromatic, defocused images acquired by the system.

The third part of the thesis explores a learning-based method that can rapidly capture 3D volumetric images of thick scattering samples using a traditional wide-field microscope. The key idea is to use a 3D generative adversarial network (GAN) based neural network to learn the mapping between the blurry low-contrast image stacks obtained using a wide-field microscope and the sharp, high-contrast image stacks obtained using a confocal microscope. After training the network with widefield-confocal stack pairs, the network can reliably and accurately reconstruct 3D volumetric images that rival confocal images in terms of lateral resolution, z-sectioning , and image contrast.