Real-time prediction based snow and rain accumulation in virtual environments

Abstract

Although weather effects have been developed thoroughly in offine renderings, their computational cost exceed current hardware to run them in real-time. Therefore, less expensive methods are developed. Still, in many real-time applications weather conditions are prepared by artists for each environment independently because developers do not want to compromise performance over simulated weather. The main reason of this is current developed algorithms use most of the graphical processing power just for weather effects in a relatively small scale scene. At the end, they become unusable in real scenario, because weather conditions are only one part of the overall system. Therefore, in current approach, all weather conditions are executed based on predictions, where its computational cost is negligible under 0.5 ms on current generation hardware. Predictions are done based on many factors such as orientation, height and position on terrain and global wind and sun direction. Appearance of weather conditions are generated from original textures of the objects then compared and re-adjusted to achieve similar values of real life scanned Bidirectional Re ectance Distribution Function (BRDF) data. All computation is done on Graphics Processing Unit (GPU) to be used with Physically Based Rendering (PBR) approach. Even though all changes are made based on prediction and textures are generated, at the end, results are comparable to weather conditions in real life.