A Survey on Methods for Omnidirectional Shadow Rendering

shadow mapping, rendering, GPU, performance, cube maps, Dual-Paraboloid mapping

This paper focuses on methods of rendering shadows cast by point light sources. The goal is to summarize advantages and disadvantages of methods based on shadow mapping. We compare the traditional approach that exploits cube maps with the Dual-Paraboloid mapping. All of the methods are implemented on the latest hardware and they exploit capabilities of current GPUs. We also implemented optimization techniques which decrease the computational time. We examine the time the methods spent in particular rendering passes and we evaluate their overall performance. Finally, we conclude the comparison with some recommendations for typical applications in which the methods of interest can be exploited. We also suggest some direction of future investigation.