Computer Graphics Laboratory ETH Zurich


Light-Field Approximation Using Basic Display Layer Primitives

N. Ranieri, S. Heinzle, P. Barnum, W. Matusik, M. Gross

SID Symposium Digest of Technical Papers 2013 (Vancouver, Canada, 19-24 May, 2013), pp. 408-411
[Abstract] [BibTeX] [PDF]


A new concept using different display layer primitives for light field approximation is presented. For each primitive, a mathematical notation for the light transport operator is defined. Based on that, a decomposition algorithm rendering a light field into a multi-layered plenoptic display is described and verified by a physical prototype.

@article {Ran13,
author = {Ranieri, Nicola and Heinzle, Simon and Barnum, Peter and Matusik, Wojciech and Gross, Markus},
title = {Light-Field Approximation Using Basic Display Layer Primitives},
journal = {SID Symposium Digest of Technical Papers},
volume = {44},
number = {1},
publisher = {Blackwell Publishing Ltd},
issn = {2168-0159},
pages = {408--411},
year = {2013},
[Download BibTeX]

Figure 1: Illustration of the steps of our light field decomposition. (1) volumetric rendering, (2) view-independent and (3) view-dependent occlusion culling, (4) view-dependent rendering on parallax type layer. Photographs are taken from our multi-layered display prototype.


We have introduced a general concept of multi-layer plenoptic displays. These display systems fuse multiple emissive and spatially modulating layers. First, we have presented a mathematical framework to analyze light transport for these displays. Second, we have described a rendering algorithm that takes as input a 3D scene and drives a given multi-planar system. Next, we have conducted an error analysis for multi-planar plenoptic displays. We then discussed practical issues of designing and building different display configurations. We demonstrate examples both in a simulation as well as on our physical prototype. Figure 1 shows all steps of our algorithm recorded on our multi-layered plenoptic display, each step decreasing the difference to the input light field. Figure 8 shows some of the simulated results that are used in our quantitative error analysis, supporting our insights about required number of layers. Though limitations imposed by current available hardware we believe our approach to suit a certain range of applications and will get more involved with upcoming technologies.

Figure 2: Simulated results used for the quantitative error analysis.


Download Paper