Research On Light Field Imaging Technology

As a computational optical sensing approach, light field imaging combines the de-sign of both optical systems and signal processing algorithms together to overcome the limitations of conventional imaging systems. A light field camera captures the four-dimensional distribution of light rays and then computationally resample them to re-construct images for required applications.This thesis includes the following work on light field imaging technology.Firstly, we theoretically illustrate the principle of light field imaging by both con-tinuous mathematical modeling and discrete signal analysis. Mathematically, light field imaging is a transform between2D and4D variables. From the view of signal process-ing, light field acquisition and reconstruction corresponds to sampling and resampling of discrete data. Considering the data redundancy of acquired light fields, we propose a novel refocusing approach based on image super-resolution reconstruction.Secondly, we design and fabricate a light field camera by inserting a microlens array in a conventional imager. On the purpose of experimental comparison, the camera includes both light field and conventional imaging systems with the same main lens. In particular, we use simulation and calculation to explore the limitation of light field resolution restricted by microlenses. We also demonstrate the mechanical design and measuring method of the light field sensor, which consists of a microlens array and a photosensor.Thirdly, we experimentally evaluate the performance of light field refocusing. A calibration method is proposed to convert the raw sensor image to4D light fields. We then compare the imaging performance of light field and conventional cameras by mea-suring their spatial resolution, depth-of-focus, modulation transfer function and point spread function, respectively. Photographic applications with light field cameras illus-trate the abilities of digital refocusing and extending depth-of-field. Fourthly, we demonstrate multi-spectral imaging applications with the light field camera. A snapshot imaging spectrometer is proposed by integrating a multi-channel spectral filter in the light field camera. We present filter design and calibration methods to avoid spectral aliasing between neighboring channels. Experimental results show that the light field imaging spectrometer is able to capture multi-spectral images in a single shot.Lastly, we present an occlusion removal algorithm in synthetic aperture imaging applications. Synthetic aperture imaging uses a camera array to capture large extent light fields and to synthesize a very large aperture, enabling seeing through partial oc-clusions. We explore the directional redundancy of light field and then propose a method to remove occlusions in synthetic images. Experimental implementation and compari-son with other algorithms illustrate that the proposed approach significantly improves the contrast and signal-to-noise ratio of synthetic aperture images.