To Predict Reality in Virtual Environments: Exploring the reliability of colour and light appearance in 3D-models
Konferensbidrag (offentliggjort, men ej förlagsutgivet), 2015
To predict how the not yet built environment is going to appear regarding light and colour is a crucial problem for architects and designers. 3D-visualization is an established design tool and used for representations of project proposals. In many cases, the aim with the visualizations is to make good-looking images sell the design. In order for software, such as 3Dsmax to become a usable planning tool of light and colour in buildings, the visualization must be trusted to show the correct appearance in accordance with the physical preconditions of reality. The complex interaction between light and objects makes the problem of lighting scenes central within computer graphics. In a previous study (2006, 2005, 2004), various problems related to the rendering of light and colour in interactive models made in 3Dsmax. Since then, the technology within colour rendering and computer graphics has moved forward. This paper discusses the problems of translating reality to its digital counterpart. New tests were performed in 3Dsmax in order to explore the usability of various rendering techniques and to explore the trustworthiness of colour and light representations. We deal with methodological issues concerning how to compare visual results between different media, mixed adaptation and arbitrary parameter setting in the software.
In the earlier studies carried out 10 years ago, a 25 m2 real room was compared to different Virtual Reality (VR) simulations. In order to get a simulation as correct as possible of the 6 fluorescent luminaires (2 ceiling armatures and 4 wall washers) the manufacturer Fagerhult’s own IES-files were used. At that time, this photometric light could only be rendered with the default Scanline renderer in 3Dsmax, which resulted in significant differences between the real and the virtual rooms. The VR-models had incorrect reflection effects between surfaces, too few colour variations and too achromatic shadows. Also they had incorrectly reproduced contrast effects for the lightest surfaces. The surface that was perceived as the whitest in the room appeared too grey.
In our new study, the original models in 3Dsmax Design 2014 were rendered with the Scanline renderer, Mental Ray and V-ray. The light planning software Dialux was used for comparison regarding the light level and distribution.
Preliminary results show improvement regarding interreflections and colour variations. However, there are still incorrectly reproduced contrast effects and the whitest areas appear too grey. The light level also differed compared to the situation in Dialux, even if the same light fixtures and IES-files where applied. Furthermore, we will discuss issues concerning the choice of correct parameters to represent the physical prerequisites of reality, such as the problem of deciding how many times a light should bounce. One dilemma is that the architects and designers have knowledge of colour and light in reality, however do not usually know how to use all features in the software. The professional visualizers know how to manipulate 3D-models to look good, but often do not regard physical correctness.