多级存储优化的大规模全局光照快速计算

针对大规模复杂场景渲染的全局光照计算量大从而影响其渲染实时性的问题,提出了一种基于多级分辨率纹理存储结构的改进体素锥追踪全局光照算法。算法首先采用基于人眼视觉特性的多级分辨率纹理存储结构高效存储光照信息,并在直接光照计算时,结合存储结果特点,然后采用混合存储的GPU硬件加速与无贡献节点丢弃的软件加速相结合方法,进一步提高光照渲染效率,最后基于多级纹理及闭合体节点自适应合并对锥波滤器进行改进,实现全局光照的高效计算。试验结果表明,改进算法减少了光照计算量,降低了系统内存占用,并取得了与经典算法相近的场景光照渲染效果,从而验证了其在大规模光照计算的有效性。

Fast calculation of large-scale global illumination optimized by multi-level storage strategy

Aiming at the problem of large amount of global illumination calculation for large-scale complex scene rendering,an improved voxel cone tracking global illumination algorithm based on multi-level resolution texture storage structure is proposed. Firstly,a multi-level texture storage structure based on human visual characteristics to efficiently store lighting information is used. And then,in the direct lighting calculation,combined with the characteristics of the storage result,a combination of GPU hardware acceleration of hybrid storage and software acceleration of discarded uncontributed nodes is used to further improve lighting rendering efficiency.Finally,the cone wave filter is improved based on the multi-level texture and the adaptive combination of closed body nodes to achieve efficient calculation of global illumination. The experimental results show that,the improved algorithm reduces the amount of lighting calculations,reduces the system memory footprint,and achieves a scene lighting rendering effect similar to the classic algorithm,thereby verifying its effectiveness in large-scale lighting calculations.