Light field representation of 3D scenes captured by event camera(s)

Abstract

A light field is the most complete representation of a 3D scene. Building a light field using standard red-green-blue (RGB) cameras has drawbacks, namely huge amount of data, limited dynamic range, and poor temporal resolution. Combining RGB and event cameras may fix these issues. This thesis explores how an event light field compares to a standard light field. The standard light field is built using standard RGB images, while the event light field is built using event aggregation images. The scene is digital, and Blender is used for rendering. The RGB rendered images are processed with an event camera simulator, which outputs events that are processed with a Python script, producing event images. The light fields are visually inspected on a light field display in the Centre for Immersive Visual Technologies (CIVIT). The result is that the event light field is the black and white outline of the standard light field. The shape of the object is visible. The event light field is much less memory-demanding than the standard one.