Unified Frequency Domain Analysis of Lightfield Cameras

Authors:
Todor Georgiev;Chintan Intwala;Sevkit Babakan;Andrew Lumsdaine
Affiliations:
Adobe Systems, San Jose, CA, 95110 and Computer Science Department, Indiana University, Bloomington, IN, 47405;Adobe Systems, San Jose, CA, 95110 and Computer Science Department, Indiana University, Bloomington, IN, 47405;Adobe Systems, San Jose, CA, 95110 and Computer Science Department, Indiana University, Bloomington, IN, 47405;Adobe Systems, San Jose, CA, 95110 and Computer Science Department, Indiana University, Bloomington, IN, 47405
Venue:
ECCV '08 Proceedings of the 10th European Conference on Computer Vision: Part III
Year:
2008

Citing 0
Cited 3

Compressive sensing of light fields

ICIP'09 Proceedings of the 16th IEEE international conference on Image processing
Computational plenoptic imaging

ACM SIGGRAPH 2012 Courses
On Plenoptic Multiplexing and Reconstruction

International Journal of Computer Vision

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper presents a theory that encompasses both "plenoptic" (microlens based) and "heterodyning" (mask based) cameras in a single frequency-domain mathematical formalism. Light-field capture has traditionally been analyzed using spatio-angular representation, with the exception of the frequency-domain "heterodyning" work. In this paper we interpret "heterodyning" as a general theory of multiplexing the radiance in the frequency domain. Using this interpretation, we derive a mathematical theory of recovering the 4D spatial and angular information from the multiplexed 2D frequency representation. The resulting method is applicable to all lightfield cameras, lens-based and mask-based. The generality of our approach suggests new designs for lightfield cameras. We present one such novel lightfield camera, based on a mask outside a conventional camera. Experimental results are presented for all cameras described.