UC Riverside

Lensless imaging system is a novel alternative to traditional camera design where the lensof the camera is replaced with another optical encoder(coded mask) that is thin, flat, and potentially inexpensive. This dissertation provides a framework of the lensless imaging model and investigates computational algorithms to extract the image and depth information of the lensless imaging measurements. The dissertation focuses on three main tracks: Lensless Imaging for 3D reconstruction Traditionally, to estimate the disparity map of the scene, we need to either take a sequence of focus stack or use binocular views with a baseline. In lensless imaging, the optical encoder placed on top of the camera sensor provides a complex and structured point spread function(PSF) that changes dramatically depending on the depth of the light source. With such property, we can retrieve the depth information from one single frame of measurements. We formulate the forward imaging model of the system such that the PSF is a continuous function of the depth so that we can estimate the depth maps more accurately on a continuous domain. We design an alternative updating algorithm based on the imaging model and jointly estimate the RGB and depth information from the measurements. Improved 3D Lensless Imaging with Learned Programmable Masks Existing methods for lensless imaging can recover the depth and intensity of the scene, but they require solving computationally-expensive inverse problems. Furthermore, existing methods struggle to recover dense scenes with large depth variations. To overcome this problem, we propose a lensless imaging system that captures a small number of measurements using different patterns on a programmable mask. We present a fast recovery algorithm that can be made parallel to recover textures on a fixed number of depth planes from multiple measurements. We also consider the mask design problem for programmable lensless cameras, and provide a design template for optimizing the mask patterns with the goal of improving depth estimation. Coded Illumination for Improved Lensless Imaging Under the space-limited circumstances, such as under-display imaging, the imaging system is often ill-conditioned and the number of sensing elements is also constrained. However, with the assistance of coded illumination patterns, we could capture more potentially uncorrelated measurements and thus improve the conditioning of the original linear system. In the proposed framework, the scene is illuminated by a sequence of simple coded illumination patterns as the lensless camera records sensor measurements. We propose a fast and low-complexity recovery algorithm that exploits the properties of the illumination patterns and system. Finally, the images reconstructed by merging the measurements from illuminated scene show significant improvement over the original camera system.