Project 1: Implementation and evaluation of Sequential Sampling for Dynamic Environment Map Illumination
Using Image Based Lighting (IBL) virtual objects can appear as if they where part of a real photograph (scene) by illuminating them with captured radiance information from the scene. Traditionally IBL techniques have mainly been applied to rendering scenes with static illumination. With the introduction of HDR video cameras it is now becoming possible to capture dynamic environment maps. Rendering objects illuminated with a dynamic environment map sequence is computationally demanding. While traditional Monte Carlo approaches treat each frame separately, recent work has considered Sequential Monte Carlo (SMC) samplers utilizing the correlation between subsequent frames and thereby enabling more efficient sampling [Ghosh et al. 2006]. However, due to the lack of real world HDR video data, the original article only considered synthetic test scenarios, and the method have yet not been put to the test in a real world scenario. In this CADICS project we have implemented the original method in a GPU based rendering system based on the NVIDIA OptiX framework. Using a state-of-art HDR video camera [Kronander et al. 2012] we have collected a set of real world dynamic environment map sequences for evaluation and compared the method to bidirectional importance sampling. Results indicate reductions in temporal flickering when using the sequential approach compared to the bidirectional approach.
Ghosh, A., Doucet, A., and Heidrich, W. 2006. Sequential sampling for dynamic environment map illumination. In Rendering Techniques, 115–126.
Kronander, J., Gustavson, S., and Unger, J. 2012. Real-time HDR video reconstruction for multi-sensor sytems. In ACM SIGGRAPH 2012 Posters.