Use this URL to cite or link to this record in EThOS:
Title: Sequential MCMC Methods For The Bayesian Inference of Patch-Based Surface Geometry
Author: Mullins, Andrew
ISNI:       0000 0001 3430 908X
Awarding Body: University of Warwick
Current Institution: University of Warwick
Date of Award: 2008
Availability of Full Text:
Access from EThOS:
This thesis is about video-based rendering, or free-viewpoint video as it sometimes known. The video-based rendering problem can be summarized as follows: after using a set of video cameras to record a dynamic scene from several different points of view, is it possible to play back that scene, but from an arbitrary point of view? First, current work in this area is reviewed and appraised. Then, this thesis makes the argument that for a general studio set-up, only two current approaches provide a realistic solution to this problem: voxel-based approaches, and polygon mesh-based approaches. That is, approaches which construct an explicit model of scene geometry to aid renderings of that scene. vVhen constructing these models of scene geometry, voxels provide a very general representation, whilst the smoothness constraint inherent in a polygon mesh provides a certain degree of robustness. This thesis makes the case for a new representation of scene geometry based on piecewise planar elements, which provides an excellent compromise between the voxel and polygon meshbased approaches. In order for such an approach to make sense, two things are necessary: some method of estimating planar elements which accurately reflect the scene geometry; and some way to produce renderings from this representation. The former is the topic of this thesis, and the problem can be broken down into three sub-problems: how to construct a suitable prior on the surface geometry; how to estimate the surface geometry given this prior; and how this surface geometry can be updated over time to reflect movement within the scene. At each stage a rigorous, Bayesian approach is taken to these problems
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available