The creation of realistic computer graphics images involves complex computations of the interaction of light and surfaces.
Global illumination algorithms compute physically based solutions but are slow by nature. For interactive rendering detached algorithms with specific approximations and simplifications are used to plausibly reproduce certain sub-sets of global illumination effects.
The goal of this project is to develop a scalable and flexible global illumination algorithm. Scalable relates to the accuracy of the solution which ranges from a plausible approximation at interactive speed to an accurate solution as obtained from non-real-time algorithms. The key aspect is that a single algorithm provides means to compute a solution as good as the computing time allows without switching between distinct, unattached algorithms. Flexible relates to the representation of the scene geometry. Traditionally, polygonal surface descriptions are used in computer graphics, but they cannot be obtained for complex real-world objects like trees. For this kind of objects, only a less accurate volumetric representation can be reconstructed from a set of photographs.
The new approach should operate on scenes with different or mixed object representations and avoid costly methods for capturing complex, real-world objects. Two fields of application will be examined:
1.) Scenes with a polygonal representation and synthetic lighting conditions using the scalable approach to achieve interactive frame rates.
2.) Scenes with natural lighting and objects, primarily volumetric reconstruction of captured trees, where we want to extend previous work to render new views with arbitrary lighting conditions and shadows.
Translucency and indirect illumination are important for a realistic appearance and are to be computed using the new approach. A comparison of the results using a volumetric and a polygonal representation can be done by using procedurally generated trees.
Call for proposal
See other projects for this call