Multiple-depth shadow maps

Abstract

Traditional shadow maps store a single depth value per cell, leading to a binary outcome by the shadow test (either lit or in shadow), and are prone to produce aliased shadow borders. We present a new approach that produces better estimates of shadow percentages and, in combination with percentage closer filtering (PCF), reduces aliasing artifacts using smaller kernel sizes. The new algorithm extends the notions of shadow maps and shadow tests to support the representation of multiple depth values per shadow map cell, as well as multi-valued shadow tests. This new approach has the potential for hardware implementation, but can also be implemented exploiting the programmable capabilities of recent graphics cards.

Rectilinear texture warping for fast adaptive shadow mapping

“…The method most commonly used for calculating shadows in interactive applications is shadow mapping [Williams 1978;Woo 1992;Brabec et al 2000;Pagot et al 2004]. Conventional shadow mapping is an easy to implement method with low computational cost, good for calculating the shadows of complex objects.…”

Sample Based Visibility for Soft Shadows using Alias‐free Shadow Maps

“…This becomes necessary due to the limited depth precision and a different discretization when rasterizing from the eye and the light. Numerous solutions have been proposed, including better adaption of the shadow map [MT04, WSP04, LSO07], antialiasing [AMB*07, PCdON04], or suggestions to lower the bias‐concerns [HN85, WE03, Woo92]. Alias‐free shadow maps exist based on either software rendering [AL04], new hardware features (still not available) [JLBM05], or depth peeling, requiring one render pass per depth layer [Arv07].…”