Detailseite
Projekt Druckansicht

Data-Driven Analysis and Synthesis of Bidirectional Texture Functions

Fachliche Zuordnung Bild- und Sprachverarbeitung, Computergraphik und Visualisierung, Human Computer Interaction, Ubiquitous und Wearable Computing
Förderung Förderung von 2006 bis 2012
Projektkennung Deutsche Forschungsgemeinschaft (DFG) - Projektnummer 17976381
 
Erstellungsjahr 2012

Zusammenfassung der Projektergebnisse

With increasing capabilities of rendering systems to accurately simulate the light transport, the creation of the necessary highly detailed scene descriptions has increasingly been recognized as a large challenge. One of the problems in this regard is the faithful reproduction of materials which give a surface its distinctive appearance, providing a lot of information to the observer about a materials characteristics. Bidirectional Texture Functions (BTFs), which are a data-driven representation obtained by taking images of real material sample from a large number of view and light combinations, have proven to be capable of reproducing a wide range of materials faithfully and preserving even subtle details. Prior to the onset of this project, there was already the necessary technology available for the capture, processing and rendering of BTFs which allowed for the accurate reproduction of a given material sample. However, in many cases the desired goal is not the accurate reproduction of a given sample, but instead the exploration of possible materials and the selection or design of a suitable new material for a given application. This project approached this problem by acquiring a large database of measured material samples. By applying suitable interpolation algorithms, these materials can then serve as a starting points for the user to design new materials by combining these samples. This enables an exploration of the space of materials by selecting materials with a set of desired perceptual properties and then creating a new one, which unites these properties. To enable this interpolation, we developed suitable representations which separate the measured BTFs into geometry and reflections properties. This allows for an interpolation independent from effects like parallax and shadows. Furthermore, we have developed a texture synthesis algorithm which can interpolate between textures. By combining these two techniques, it is possible to interpolate between two measured BTFs to obtain materials which are perceptually in between two input samples, both in regard to the reflection properties as well as the geometric structure. There remain several questions for further investigation. So far, we are only able to interpolate between two materials, which limits the possibilities of a designer and impedes his ability to explore the space of materials along several perceptual traits. Extending our techniques for several materials and characterizing the space spanned by these materials perceptually remain topics for further research. Similarly, performance improvements both in regard to computation time and storage space requirements and a perceptual evaluation for which material classes the developed techniques are already sufficient and for which different approaches are needed remain avenues of further research.

Projektbezogene Publikationen (Auswahl)

 
 

Zusatzinformationen

Textvergrößerung und Kontrastanpassung