Nousias, Sotirios D.; (2021) From Calibration to Large-Scale Structure from Motion with Light Fields. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text 2021___Nousias___PhD_Thesis_final.pdf - Accepted Version Download (87MB) | Preview

Abstract

Classic pinhole cameras project the multi-dimensional information of the light flowing through a scene onto a single 2D snapshot. This projection limits the information that can be reconstructed from the 2D acquisition. Plenoptic (or light field) cameras, on the other hand, capture a 4D slice of the plenoptic function, termed the “light field”. These cameras provide both spatial and angular information on the light flowing through a scene; multiple views are captured in a single photographic exposure facilitating various applications. This thesis is concerned with the modelling of light field (or plenoptic) cameras and the development of structure from motion pipelines using such cameras. Specifically, we develop a geometric model for a multi-focus plenoptic camera, followed by a complete pipeline for the calibration of the suggested model. Given a calibrated light field camera, we then remap the captured light field to a grid of pinhole images. We use these images to obtain metric 3D reconstruction through a novel framework for structure from motion with light fields. Finally, we suggest a linear and efficient approach for absolute pose estimation for light fields.

Download activity - last month

Export as XMLCSVJSON

Download activity - last 12 months

Export as JSONCSVXML

Downloads by country - last 12 months

Export as JSONXMLCSV

Archive Staff Only

View Item