Scale-invariant analysis of light fields using Lisad spaces
Event details
| Date | 03.11.2014 |
| Hour | 16:15 › 17:45 |
| Speaker | Dr Ivana Tošić, Ricoh Innovations |
| Location | |
| Category | Conferences - Seminars |
Recent development of hand-held plenoptic cameras has brought light field acquisition into many practical and low-cost imaging applications. The acquired light fields capture both spatial and angular information of light rays passing through the camera aperture. As such, they intrinsically capture 3D information of the imaged scene. However, 3D scene reconstruction and analysis from light fields still remains a challenge due to their high dimensionality and their particular structure. To address this problem, we propose a new theory for scale-invariant 3D analysis of light fields via the construction of Light field scale-and-depth (Lisad) spaces, which are parametrized both in terms of scale of objects recorded by a light field and in terms of objects' depth. I will show two applications of Lisad spaces: dense depth estimation and 3D keypoint detection from light fields. In both cases, our algorithms based on Lisad spaces outperform their respective prior art counterparts. Moreover, their implementation requires only local processing, which shows great potential for integration of Lisad-based methods into future computer vision architectures for plenoptic cameras.
This is joint work with Kathrin Berkner at Ricoh Innovations, Corp.
Bio:Ivana Tošić received a Ph.D. degree in computer and communication sciences from the Swiss Federal Institute of Technology (EPFL), Switzerland in 2009, and a Dipl.Ing. degree in telecommunications from the University of Niš, Serbia in 2003. From 2009 to 2011, she was a postdoctoral researcher at the Redwood Center for Theoretical Neuroscience, University of California at Berkeley, where she studied computational mechanisms of depth perception from binocular vision. Since 2011, Dr. Tošić has been a member of research staff at Ricoh Innovations, Corp., Menlo Park, California, working in the computational optics and visual processing group. Her research interests lie in the intersection of image processing and computational neuroscience domains and include binocular vision, image and 3-D scene representation, depth perception, representation and coding of the plenoptic function, and computational photography.
This is joint work with Kathrin Berkner at Ricoh Innovations, Corp.
Bio:Ivana Tošić received a Ph.D. degree in computer and communication sciences from the Swiss Federal Institute of Technology (EPFL), Switzerland in 2009, and a Dipl.Ing. degree in telecommunications from the University of Niš, Serbia in 2003. From 2009 to 2011, she was a postdoctoral researcher at the Redwood Center for Theoretical Neuroscience, University of California at Berkeley, where she studied computational mechanisms of depth perception from binocular vision. Since 2011, Dr. Tošić has been a member of research staff at Ricoh Innovations, Corp., Menlo Park, California, working in the computational optics and visual processing group. Her research interests lie in the intersection of image processing and computational neuroscience domains and include binocular vision, image and 3-D scene representation, depth perception, representation and coding of the plenoptic function, and computational photography.
Practical information
- Informed public
- Free
- This event is internal
Organizer
- STI-IEL-LTS4
Contact
- Prof. Pascal Frossard