Visual acceptance evaluation of soft shadow algorithms for virtual TV studios

Shadows in computer graphics are an important rendering aspect for spatial objects. For realtime computer applications such as games, it is essential to represent shadows as accurate as possible. Also, various tv stations work with virtual studio systems instead of real studio sets. Especially for those systems, a realistic impression of the rendered and mixed scene is important. One challenge, hence, is the creation of a natural shadow impression. This paper presents the results of an empirical study to compare the performance and quality of dierent shadow mapping methods. For this test, a prototype studio renderer was developed. A percentage closer lter (pcf) with a number of specic resolutions is used to minimize the aliasing issue. More advanced algorithms which generate smooth shadows like the percentage closer soft shadow (pcss) method as well as the variance shadow maps (vsm) method are analysed. Dierent open source apis are used to develop the virtual studio renderer, giving the benet of permanent enhancement. The Ogre 3D graphic engine is used to implement the rendering system, beneting from various functions and plugins. The transmission of the tracking data is accomplished with the vrpn server/ client and the Intersense api. The dierent shadow algorithms are compared in a virtual studio environment which also casts real shadows and thus gives a chance for a direct comparison throughout the empirical user study. The performance is measured in frames per second.