Research On Real-Time Soft Shadow Techniques In Large-scale Virtual Environments

Shadow effect is very important for the realism of virtual environment. This is because shadows provide direct visual information which helps us to understand the spatial relationships of objects in the scene. In the field of computer graphics, shadow rendering is a very hot topic with great research value. Researchers have proposed a lot of shadow rendering algorithms. However, calculating soft shadows which have full compliance with the laws of physics in real-time is still very difficult and the researchers were only able to use some ideal model to approximate the real shadows.In large-scale virtual environments, the area of the scene is large, and the number of the scene objects is big. Also, their distribution and occlusion relationships are quite complicated. In rendering such a virtual scene, the view frustum needs to cover a large range of volume and this increases the complexity of shadow effect rendering. Most of the current shadow rendering techniques are not for this particular application environment, and can't solve the balance problem between real-time request and soft shadow rendering quality well. This study focuses on scene space partitioning, shadow mapping with multiple shadow maps and adaptive filtering to maintain the real-time shadow rendering and improve the fidelity of soft shadows.The work of this paper handles the existing problem of soft shadow rendering in large-scale complex virtual scenes. The method proposed by this paper splits the whole shadow rendering problem into several simpler parts, solves them one by one adaptively and then does the integration work to get the final result. The experiment results show that our method can meet real-time requirements, while achieving the realistic soft shadow effect.The main research content and achievements of this paper are as follows: (1) Present a new methodological framework named parallel-split percentage closer soft shadow appropriate for rendering soft shadows in large-scale virtual environments. This method divides the scene space into a series of smaller partitions along the z-axis, draws an independent shadow map on each partition, which has a smaller resolution when is farther from the viewer and then calculates soft shadows of corresponding scene partition based on the parallel occlusion assumption. This LOD based approach fully enhances the utilization of the shadow map resolution and balances the shadow rendering speed and quality well.(2) In-depth study the image-space-based shadow mapping techniques, analyze the aliasing error of them, and replace the traditional shadow map in PSPCSS with variance shadow map to improve the quality of soft shadow based on shadow reconstruction function analysis.(3) According to the"divide and conquer"idea, we modify the process of blocker research and soft shadow calculation in PSPCSS from two perspectives, the global shadow map layer division and the local filter kernel subdivision. This results in further improvement of the PSPCSS efficiency and shadow rendering quality.