Deep Shadow Maps (DSM)

Regular shadow maps store a single depth value per pixel, to determine whether a certain point in the scene is either illuminated by the light source or completely in shadow. Deep shadow maps instead store one function per pixel to compute a partial visibility value for every possible depth in its corresponding frustum in the scene. This makes it possible to generate fast, shadow-map based shadows for scenes with semi-transparent volumes like fog or smoke, or layers of semitransparent objects, like fur and hair. Because the stored data is a 2D array of functions and not a densely sampled 3D volume, the representation requires only reasonable amounts of storage on disk and in memory.

• Not currently implemented in Aqsis 1.2. Support is planned for 1.3 and work is in progress.