[Back]

F. Kreuzer:
"Using Perception-Based Filtering to Hide Shadow Artifacts";
Supervisor: M. Wimmer; Institut für Computergraphik und Algorithmen, 2017; final examination: 2017-08-24.

Shadows are an indispensable aid for understanding spatial relations of objects in natural scenes, which is why they are very important for real-time rendering applications. Combining filtering techniques with shadow mapping is a common tool to simulate visually pleasing shadows in interactive applications. A positive effect of such approaches is that filtering blurs aliasing artifacts caused by sampling the discretized geometric data stored in the shadow map, thereby improving the visual quality of the shadow. The goal of this thesis is to exploit common filtering algorithms in order to find a function of blur radius and shadow-map sampling frequency, which allows for optimized computational performance while mostly preserving the visual quality of the shadow. In the course of this work, we investigate how shadow artifacts arise and how to hide them. We set up and execute a user study to find the optimal relation between the shadow-map sampling frequency and the filter radius. From the results of the user study, we derive a formula and develop an algorithm that can be incorporated into existing shadow-mapping algorithms. We evaluate our results by applying the algorithm to a custom-made rendering framework and observe an increase in processing speeds.

http://publik.tuwien.ac.at/files/publik_261866.pdf