Multi-resolution screen-space ambient occlusion

Abstract

We present a new screen-space ambient occlusion algorithm (SSAO), which we call Multi-Resolution Screen-Space Ambient Occlusion (MSSAO). The method computes ambient occlusion by combining occlusion values using multiple mipmap levels of a g-buffer. Our method is based on the observation that occlusion due to further-away occluders are lower frequency and thus can be computed in coarser resolutions. On the other hand, nearby occluders cause higher-frequency occlusion that must be computed in finer resolutions. Our method samples occluders in screen space (similar to [Shanmugam and Arikan 2007]) and combines occlusion values across multiple image resolutions. Compared with two state-of the-art SSAO methods [Filion and McNaughton 2008; Bavoil et al. 2008], results from our algorithm are closer to ray-traced results, even when our algorithm is running at comparable or higher frame rates. Because no random sampling is used, and occlusion values in all levels are combined using edge-preserving blending, our results are free of noise or excessive blurring which are common in other SSAO methods. Copyright © 2010 by the Association for Computing Machinery, Inc.