Scheme: University Research Fellowship
Organisation: Heriot-Watt University
Dates: Oct 2014-Sep 2019
Summary: Imagine sports fanatics being able to interactively modify their individually personalised camera view of an actual live football broadcast. Imagine security personnel scouring through large amounts of archived visual data recorded in a bustling airport terminal, with interactive view control to look behind obstacles. Envision a primatologist with a tool to record the climbing movements of monkeys in an instrumented lab-forest environment, and later being able to observe different individuals, following their paths as they weave through branches. The viability of acquiring plenoptic functions is a first step towards such interactive exploration of dynamic, real environments to an extent that is only currently possible in virtual environments. One way to achieve this would be to tabulate all rays (eg. in a football stadium) and record the amount of light they carry. Then, generating a unique view reduces to looking up the value in the table for each ray that should arrive at the view location, along the gaze direction. Clearly this is infeasible.
My research addresses the problem of carefully selecting a subset of views that will yield plausible reconstructions from hypothetical views, and an investigation of the trade-off between error in the hypothetical views and the number and nature of cameras used to capture the original views.
Scheme: Newton International Fellowships
Organisation: University College London
Dates: Jan 2011-Dec 2012
Summary: This project summary is not available for publication.