Hard shadows:
At a broad level, prior techniques to alleviate
aliasing artifacts using rasterization methods are based on shadow
maps [Williams 1978] and shadow volumes [Crow 1977]. Some
hybrid approaches have been proposed that combine shadow map-
ping and volumes [McCool 2000; Chan and Durand 2004] that can
improve shadow volume performance and allow interactive high-
quality shadows on simple scenes. Most current interactive applica-
tions use variants of shadow mapping, but may suffer from aliasing
problems. Many practical algorithms have been proposed to allevi-
ate perspective aliasing [Stamminger and Drettakis 2002; Wimmer
et al. 2004; Lloyd 2007] as well as projective aliasing [Lefohn et al.
2007]. Other shadow mapping algorithms can eliminate blocking
artifacts [Aila and Laine 2004; Johnson et al. 2005] by implement-
ing a rasterizer that can process arbitrary samples on the image
plane.
Soft shadows:
In general, soft shadows can be implemented by
sample-based methods such as using averaging visibility from mul-
tiple shadow maps to calculate visibility or ray tracing. Both these
methods can be slow, so many approaches have been developed to
generate plausible soft shadows with methods such as post-filtering
shadow maps or special camera models [Mo et al. 2007], which
produce correct results only for simple scenes. More accurate ap-
proaches evaluate the light source visibility from the image samples
by back-projection [Assarsson and Akenine-Möller 2003; Schwarz
and Stamminger 2007; Sintorn et al. 2008; Bavoil et al. 2008] or by
generating shadows from environment lighting [Annen et al. 2008].
Techniques using irregular z-buffering have also been extended for
soft shadows on a proposed new architecture [Johnson et al. 2009]. | 
Linear Mode
