Efficient Measurement and Compression of multi-spectral Bidirectional Texture Functions

Abstract

Details

Description

In the future, easy capture of spectral resolved material data will stimulate new techniques for physically correct simulation of objects under real illumination conditions. This will impact the areas of cultural heritage, material sciences, textile design and e-commerce. Especially in e-commerce, correctness of simulated colors of products and their display on current devices is crucial and cannot be guaranteed now.
In the meantime, the acquisition of optical material parameters is well advanced. First reflectance measurements were performed using photometric measurements and gonioreflectometer setups. Nowadays, efficient setups based on CCD cameras exist, where one photograph delivers many reflectance measurement values. Up to now, these measurements are performed with RGB sensors, because the low prices for such cameras led to cheap measurement setups. But, since different light sources might have very different spectra, the colors of the measured materials cannot be reproduced under arbitrary illumination. Furthermore, RGB measurements contain systematic chromatic errors since the metamerism properties of the RGB filters used do not correspond with the properties of the human eye good enough.
In modern rendering systems all light simulation is done on a complete spectral basis. For special materials like car paint, the reflection behaviour is captured in a few directions using gonioreflectometers. Under certain assumptions, this is sufficient for the fitting of analytical models to the data. But such setups are insufficient for anisotropic materials or for materials with strong mesostructure and/or spatial variations. For these materials there are currently no measurement setups at hand. Similar setups like the RGB based ones are impractical for spectral measurements because of the high costs of cameras and light sources needed for spectral measurements.
In this project we plan to combine RGB and spectral measurement methods to come up with an efficient and pratical measurement setup for spectral BTFs. Furthermore, algorithm for analysis, compression and efficient rendering for such RGB-spectral-combined data will be investigated.

You can find a description of our lab here.

News

  • 03-12-2010: 1000W QTH lamp integrated into measurement setup
  • 03-08-2010: Several multispectral textures captured
  • 02-20-2010: Our paper "Groundtruth Data for Multispectral Bidirectional Texture Functions" was accepted at CGIV 2010
  • 05-31-2010: Our paper "Spectralization: Reconstruction spectra from sparse data" was accepted at EGSR 2010
  • 08-15-2011: Our paper "Practical Spectral Characterization of Trichromatic Cameras" was conditionally accepted as SIGGRAPH ASIA 2011
  • 03-29-2012: Added spectral textures
  • 04-04-2012: Uploaded first spectral BTFs

Publications

 
In proceedings of 12th International Conference on Computer Graphics Theory and Applications (VISIGRAPP), 2017
 
In proceedings of Eurographics Workshop on Material Appearance Modeling, pages 15-20, The Eurographics Association, 2015
 
In: Journal of WSCG (June 2014), 22:2(83-90)
 
In: ACM Trans. Graph. (Dec. 2011), 30:6(170:1-170:10)
 
In proceedings of CGIV 2010, Society for Imaging Science and Technology, Joensuu, Finland, pages 326-330, June 2010
 
Jason Lawrence and Marc Stamminger (Editors)
In proceedings of SR '10 Rendering Techniques, pages 1347-1354, Eurographics Association, June 2010
 
Daniel Lyssi
In proceedings of CESCG, Apr. 2009

Multispectral BTF data

Published in the BTF database in an own category (direct link).

Multispectral textures

Some multispectral textures we captured during the project. The images are in OpenEXR format with the channel names being the respective wavelength.

Multispectral Environment Maps

As part of the project we captured a full multispectral environment map of a room at the University of Bonn which has a very complex illumination consisting of large neon lamps and small halogen spotlights.

The environment map was captured using a Photometrics CoolSnap 4k monochrome camera and a CRi VariSpec VS10 liquid crystal tuneable filter. It was sampled from 410 to 720 nm wavelength in 10nm steps. High dynamic range was generated using 11 different exposure times.

The multispectral image can be downloaded undefinedhere. It is in OpenEXR format with the channels named after the respective wavelength.