Although the application is specific, the techniques I will outline are quite general. Accurate recovery of geometry requires robust methods for determining camera position across multiple views. Recovering structured geometry is essential for being able to interact with a scene - for example, to move objects around, and to fill in previously occluded detail. And recovery of bi-directional reflectance distribution functions (BRDFs) is necessary if we wish to change the global illumination in an environment. All of these are essential for reconstructing crime scenes, if we are to have any faith that the reconstruction is anything like the real scene.
This work is part of our group's REVEAL project, Reconstruction of Environments with Accurate Lighting, in which we are attempting to solve some fundamental research problems, and also looking at their application, in collaboration with the Greater Manchester Police. The project is funded by the UK's EPSRC, grant number GR/M14531.
ABOUT THE SPEAKER...
Roger Hubbold is a Senior Lecturer in Computer Science at the University of Manchester. He heads the Advanced Interfaces Group, researching software systems for large-scale collaborative virtual environments, parallel algorithms for high-quality image synthesis, and scientific visualisation, supported by grants from the UK research councils and the European Commission, and by major international companies. He has been named investigator on 19 research grants (Principal Investigator on 10), and has published over 60 refereed papers and books in his field. He is a Fellow and former Chairman of Eurographics, and an affiliate member of ACM and IEEE Computer Society.