ABSTRACT
Screen-space volume generation methods provide very fast and guaranteed response times compared to geometry-based techniques but suffer from viewdependency. More specifically, any technique that is performed entirely in screen space (as in deferred shading) considers only geometry that has been rendered into the depth buffer and thus has the following strong limitations: First, it ignores geometry located outside the field of view. Second, it ignores geometry that is inside the view frustum but occluded by other objects. Yet these geometry parts may have a significant influence on the desired final result (see our indirect illumination case study in this article).
