Stenciled Screen-Space Ambient Occlusion
Rennai Mao, Fuyan Liu · 2024
Screen-space Ambient Occlusion (AO) is to calculate the exposure of each pixel in the screen to ambient lighting. Since ambient occlusion occurs only in certain regions of the image (corners, cracks, etc.), the computation of non-occluded surfaces is often considered redundant and unfavorable. This paper first explored the relationship between artifacts and pre-pass sampling numbers in VAO++ [1]. By increasing the number of sampling points during low-resolution AO calculation, we solved the phenomenon of artifacts. Compared with the original ‘careful’ solution, it better solves artifacts and saves a small amount of sampling calculation in non-AO regions. We also propose a new method for culling non-AO regions through temporal coherence and propose a solution to the impact of dynamic objects and the problem of error accumulation. The results show that Our algorithm effectively improves the computational speed in some cases without significant loss of quality, and the improvement is more obvious when the screen resolution is higher and the number of sampling points is larger.