How to Build a Good Light Field Camera?
jueves octubre 16, 2014 - 07:30:00
Marvell Semiconductor Inc.
5488 Marvell Ln
Santa Clara,CA 95054
Goldfish Room, Building 2, First Floor
Chia-Kai Liang, PhD
Architect, Computational Photography
Obtaining a light field from real world the can enable many applications, such as post-capture camera control and scene reconstruction. While many light field capture systems have been proposed and built in the lab, the microlens/lenslet-based approach is still the most mature one and already available on the market. However, it is often stated that there is a fundamental tradeoff between spatial and angular resolution of lenslet light field cameras, but there has been limited understanding of this tradeoff theoretically or numerically. This talk will describe a light transport framework for understanding the fundamental limits of light field camera resolution.
The new framework includes many factors that were not included in existing models. For example, we show that the 4D spatial-angular sensitivity profile of the photosensor must be considered in the performance analysis. The new framework is used to make many predictions that are closer to the observations and contradict existing models. The talk will discuss how the new framework should be considered in designing the light field cameras or processing algorithms.
Chia-Kai Liang received the B.S. and Ph.D. in 2004 and 2009 respectively, both from National Taiwan University. He joined Lytro at 2010 and is currently the architect of the computational photography group, where he developed several core light field and image processing algorithms used in the Lytro first gen camera and the Lytro ILLUM camera. His research focus is on computational photography, low-level computer vision and image processing algorithms and pipeline architecture design. Chia-Kai has published 10 journal papers and is a named inventor of 8 US patents. He received the IEEE CSVT Transactions Best Paper Award in 2008 and a member of IEEE and ACM SIGGRAPH.
More information of Chia-Kai Liang is available at http://chiakailiang.org