Metameric Light Fields

Prithvi Kohli, David R. Wal­ton, Rafael Kuffn­er dos Anjos, Antho­ny Steed and Tobias Ritschel

To be pre­sent­ed as a poster at IEEEVR 2022

[Poster] | [Short Paper] | [Teas­er Video] | [Project Page]

Ven­tral metamers, pairs of images which may dif­fer sub­stan­tial­ly in the periph­ery, but are per­cep­tu­al­ly iden­ti­cal, offer excit­ing new pos­si­bil­i­ties in foveat­ed ren­der­ing and image com­pres­sion, as well as offer­ing insights into the human visu­al sys­tem. How­ev­er, exist­ing lit­er­a­ture has main­ly focused on cre­at­ing metamers of sta­t­ic images. In this work, we devel­op a method for cre­at­ing sequences of metamer­ic frames, videos or light fields, with enforced con­sis­ten­cy along the tem­po­ral, or angu­lar, dimen­sion. This great­ly expands the poten­tial appli­ca­tions for these metamers, and expand­ing metamers along the third dimen­sion offers fur­ther new poten­tial for compression.

Comparison: Metameric Light Fields vs Baseline Methods

Ground Truth
Metamer­ic Light Field (ours)
2D Only
2D with Coher­ent Noise (Kass and Pesare)
3D Pyra­mid (naive)

Comparison: Ground Truth vs Metameric Light Fields

Compression Examples

Acknowledgements

This work was fund­ed by the EPSRC/UKRI project EP/T01346X/1.