传统软阴影算法在GPGPU构架上的优化技术研究

Study of Traditional Soft Shadow Algorithm Optimization Technology in the GPGPU Framework

针对百分比靠近软阴影PCSS算法存在阴影粉刺和偏离的问题,以及图形渲染实时性的要求,提出一种改进的软阴影算法.该方法在PCSS算法的基础上,利用深度梯度消除阴影粉刺和偏离现象,使用方差阴影映射VSM算法进行过滤,利用GPU的并行计算能力加速生成区域求和表以实现对阴影图预先滤波.实验结果表明,该方法能够得到更高的软阴影质量,并且可以获得很好的加速比,实时性明显提高.

Aiming at the realistic need of shadow rendering, on the basis of the PCSS algorithm, this paper puts forward a soft shadow generation algorithm based on GPGPU. It samples a region by poisson disk pattern instead of regular sampling mode, uses depth gradient to deal with the phenomenon of self-shadowing, uses the VSM algorithm to filter the shadow, and uses the summed area table to filter the shadow map dynamically, which can improve the shadow quality and rendering efficiency in a certain extent.