RGL
Realistic Graphics Lab
EPFL Logo

Wide Gamut Spectral Upsampling with Fluorescence

In Computer Graphics Forum (Proceedings of Eurographics Symposium on Rendering 2019)

A spec­trally rendered scene with stand­ard RGB to spec­trum up­sampling (left) can­not achieve re­flect­ance spec­tra with ar­bit­rary col­our sat­ur­a­tion and bright­ness due to en­ergy con­ser­va­tion con­straints. Our tech­nique (centre) con­verts bright and deeply sat­ur­ated in­put RGB col­ours to reg­u­lar re­flect­ance spec­tra and en­hances them with a fluor­es­cent com­pon­ent, re­pro­du­cing col­ours from a con­sid­er­ably wider gamut. The in­put col­ours were spe­cified in ACEScg—for in­stance, the top ring of the swim­ming pool has col­or (0.9, 0, 0.9) in ACEScg. To visu­al­ise col­ours of such high sat­ur­a­tion, we in­ter­pret the rec2020 data of the rendered im­ages as sR­GB for dis­play, and we also provide a dis­play-in­de­pend­ent dif­fer­ence im­age in CIE74 Delta E on the right (stopped down by 7ev, i.e. a sat­ur­ated pixel value of 1.0 cor­res­ponds to Delta E = 128). Sim­il­arly, the squares to the left visu­al­ise the ACEScg in­put col­ours re­in­ter­preted as sR­GB.

Abstract

Phys­ic­ally based spec­tral ren­der­ing has be­come in­creas­ingly im­port­ant in re­cent years. However, as­set tex­tures in such sys­tems are usu­ally still drawn or ac­quired as RGB tri­s­tim­u­lus val­ues. While a num­ber of RGB to spec­trum up­sampling tech­niques are avail­able, none of them sup­port up­sampling of all col­ours in the full spec­tral locus, as it is in­trins­ic­ally big­ger than the gamut of phys­ic­ally val­id re­flect­ance spec­tra. But with dis­play tech­no­logy mov­ing to in­creas­ingly wider gamuts, the abil­ity to achieve highly sat­ur­ated col­ours be­comes an in­creas­ingly im­port­ant fea­ture. Real ma­ter­i­als usu­ally ex­hib­it smooth re­flect­ance spec­tra, while com­pu­ta­tion­ally gen­er­ated spec­tra be­come more blocky as they rep­res­ent in­creas­ingly bright and sat­ur­ated col­ours. In print me­dia, plastic or tex­tile design, fluor­es­cent dyes are ad­ded to ex­tend the bound­ar­ies of the gamut of re­flect­ance spec­tra. We fol­low the same ap­proach for ren­der­ing: we provide a meth­od which, giv­en an in­put RGB tri­s­tim­u­lus value, auto­mat­ic­ally provides a mix­ture of a reg­u­lar, smooth re­flect­ance spec­trum plus a fluor­es­cent part. For highly sat­ur­ated in­put col­ours, the com­bin­a­tion yields an im­proved re­con­struc­tion com­pared to what would be pos­sible re­ly­ing on a re­flect­ance spec­trum alone. At the core of our tech­nique is a simple para­met­ric spec­tral mod­el for re­flect­ance, ex­cit­a­tion, and emis­sion that al­lows for com­pact stor­age and is com­pat­ible with tex­ture map­ping. The mod­el can then be used as a fluor­es­cent dif­fuse com­pon­ent in an ex­ist­ing more com­plex BRDF mod­el. We also provide im­port­ance sampling routines for prac­tic­al ap­plic­a­tion in a path tracer.

Text citation

Alisa Jung, Alexander Wilkie, Johannes Hanika, Wenzel Jakob, and Carsten Dachsbacher. 2019. Wide Gamut Spectral Upsampling with Fluorescence. In Computer Graphics Forum (Proceedings of Eurographics Symposium on Rendering) 38(4).

BibTeX
@article{Jung2019Wide,
    author = {Alisa Jung and Alexander Wilkie and Johannes Hanika and Wenzel Jakob and Carsten Dachsbacher},
    title = {Wide Gamut Spectral Upsampling with Fluorescence},
    journal = {Computer Graphics Forum (Proceedings of Eurographics Symposium on Rendering)},
    volume = {38},
    number = {4},
    year = {2019},
    month = oct,
    doi = {10.1111/cgf.13773}
}