Example PhD

Lightfield Rendering

Supervisor: Professor N.J. Avis

Keywords: Lightfield Rendering, Lumigraph, GPUs, Visualization

With the recent advances in graphics cards, powerful GPUs are now available to each user of a PC at very modest cost. However, many datasets are still too large to transmit over the network for local display, especially when there are large numbers of simultaneous users. The rendering of such datasets needs to be carried out centrally, and then the resulting views distributed to the remote users.

Current GPUs are designed to process a single view for a single user; they are not designed to scale to support multiple, independent views in an efficient manner. This research programme to investigate alternative methods that can supply rendered views to large numbers (e.g. 1000) of simultaneous users.

Light field rendering supports rapid rendering of multiple views, but has so far not been used with multiple users, nor scaled to render multiple simultaneous views. The proposed work will investigate this, using initially a traditional shared memory multiprocessor architecture to allow concentration of effort on solving the challenges presented by the multi-view application. These solutions will then be ported to create a solution that works on common desktop PCs in a Grid environment. This will result in a system which will be able to utilize many slower machines (with varying amounts of memory) to support many simultaneous users through distributing the workload and the dataset.

Light field rendering is currently designed to work with a large pre-processing stage, with the dataset not being updated after that point, with the camera position also being severely limited. This research programme will investigate methods to free up the camera position, and to enable later updates to the dataset without another large pre-processing stage being required.

Key Skills/Background: Open to Computing Graduates and Postgraduates

Contact: Professor N.J. Avis to discuss this research topic.