View-dependent shadow maps

Shadows play an important role in human perception of the world because they provide us with the information about relative position of objects and their size in three dimensional space. Therefore they are indespensable in rendering realistic images in computer graphics. Precise computation of the shadows in real time is, however, a very expensive operation. In order to compensate for this, several alternative approaches were developed at cost of certain reduction of visual quality. Among these approaches, the probably most frequently used method is called shadow mapping. However, its speed and simplicity is paid back by several drawbacks. Probably the most significant one is dependency on limited image resolution of graphic cards which causes discretization errors. These errors lead to inaccurate rendering of the shadow boundaries which is known as aliasing. This paper presents an algorithm which uses the original shadow mapping method, and by utilizing the capabilities of modern graphic cards removes the limiting dependency and renders correct shadows in any resolution.