Physically Plausible Skylight Rendering

The aim of our projcet is to develop physically plausible analytical models of skylight radiance and aerial perspective that can be used for the puroposes of predictive photorealistic rendering. Although some of the existing techniques in this area already capable of reaching a high level of visual realism, none of them is capable of reliably predicting the appearance of outdoor scenes. While total accuracy is not essential for artistic applications - such as movie production work - other areas such as architectural illumination simulations and virtual prototyping applications depend an a correct evaluation of a number of subtle atmospheric effects. Forr example, the appearance of modern high-rise buildings with all-glass facades can change significantly if skylight polarization - which under certain conditions can reich up to 90% in some areas of the skydome - is taken into account. The importance of considering polarization also extends to certain industrial prototyping visualizations, e.g. in the car industry, where polarization evaluation can play a role in the correct assessment of unwanted interreflections between windshield, dashboard and other parts of the vehicle bodywork. Prediction of plausible spectral radiance distributions for variying atmospheric conditions is also highly important if any meaningful colourimetric accuracy for the prediction of object appearance is desired. In particular, computing the appearance of fluorescent materials under daylight illumination is impossible with existing skylight models, since they do not provide data an the near ultraviolet portion of the spectrum. A large number of modern dyes and pigments are at least partially fluorescent, which makes their appearance all but intraciable to renderers which do not take this feature into account. Our project work will approach this problem area from two distinct directions. One will be the derivation of suitable analytical models of the various properties of skylight radiance (e.g. the proper spectral radiance distribution and polarization state) as well as development of a new treatment of aerial perspective, and the other will focus an the improvement of those modern bidirectional photorealistic rendering techniques which currently cannot be used efficiently for outdoor scenes. The latter point is particularly important as improvements in the accuracy of skylight and aerial perspective models cannot be fully exploited as Ion-, as some of lhe best modern rendering algorithms have problems using them.