Lens Array Based Techniques for 3D Scene Capture and Display

摘要

This thesis discusses the use of lens arrays for both capture and display of 3D visual scenes while utilizing the ray optics formalism for modeling the propagation of light. In 3D capture, the use of lens arrays brings the concepts of focused and defocused plenoptic cameras, and in 3D display, the same optical technology brings the integral imaging (InI) and super multiview (SMV) visualization techniques. Plenoptic cameras combine a lens array with a single sensor in order to capture the light field (LF) emanated by a scene compactly and in a single shot. In the thesis, comparative analysis of focused and defocused plenoptic cameras is carried out in terms of LF sampling and spatio-angular resolution trade-offs. An algorithm for simulating ground-truth plenoptic image data for the case of defocused plenoptic camera is developed and implemented. It models the process of plenoptic capture and makes use of the notion of densely sampled light field (DSLF) for the sake of efficient and reliable data processing. 3D displays aim at visualising 3D scenes as accurate as possible, thus providing natural viewing experience. They are characterised and compared by their ability to correctly reproduce 3D visual cues, such as vergence, binocular disparity, accommodation and motion parallax. Design-wise, lens array based 3D display techniques provide simple yet effective way to correctly deliver all these cues, which makes them attractive in several 3D display applications. The thesis studies SMV and InI techniques in terms of depth perception and resolution trade-offs. Based on the theoretical analysis, a prototype SMV head-up display (HUD) system is developed. It demonstrates a compact and affordable solution for the virtual image presentation HUD problem. The experiments and analyses carried out on the prototype verify the SMV display capabilities for the targeted HUD application.