Abstract
An adaptive sampling-based atmospheric scattering and volumetric light framework for flight simulator (FS) is proposed to enhance the immersion and realism in real-time. The framework comprises epipolar sampling (ES), visible factor culling (VFC), interactive participating media density estimating (IPMDE). The main process of proposed architecture is as follows: the scene is divided into two levels according to the distance from the camera. In the high-level pipeline, the layer close to the camera, more samples, and smaller sampling step size is used to improve image quality. Further, the IPMDE method is designed to enhance realism by achieving interactivity with the participating media and multiple light sources. Further optimization is performed by the lookup table and 3D volumetric textures, by which we can compute the density of participating media and the scattering coefficient in parallel. In the low-level pipeline, when the samples are far away from the camera, its influence on the final output is also reduced, which results in fewer samples and a bigger sampling step size. The improved ES method further reduces the number of samples involved in ray marching using the features of epipolar geometry. It then generates global light effects and shadows of distant terrain. The VFC method uses an acceleration structure to quickly find the lit segments which eliminate samples blocked by obstacles. The experimental results demonstrate our architecture achieves a better sense of reality in real-time and is very suitable for FS.
Publisher
Public Library of Science (PLoS)