Light space parallel-split shadow maps based on programmable GPUs

Parallel-split shadow maps (PSSM) suffers from redundant rendering that unnecessarily renders same objects to multiple shadow map textures, when the light's direction and the camera's view direction are not perpendicular. To solve this problem, this paper presents a light space parallel-split shadow maps algorithm (LiSPSSM). The algorithm first divides the scene into certain layers that do not intersect each other by light's frustums in light space, and generates one shadow map for each layer to ensure that any shadow map does not have the same sample points. Then a faster synthesizing scene shadow method, which avoids operations of pixel layer determination, is given to improve GPU's efficiency. The results show that LiSPSSM effectively addresses the redundant rendering problem in PSSM and improves the efficiency and quality of real-time shadow rendering.