Sub-pixel shadow mapping

Abstract

Figure 1: Comparisons of standard Shadow Mapping with Sub-Pixel Shadow Mapping (SPSM). On the left, the YeahRight model (≈ 150 000 polys) using a 1K × 1K shadow map rendered in 11 ms with SPSM. On the right, a large City model (≈ 2 300 000 polys) rendered using a single 2K × 2K shadow map in 41.6 ms. SPSM brings sub-pixel accuracy to shadow maps, avoiding aliasing artifacts while preserving the rendering speed. AbstractThe limited resolution of shadow maps (SM) may result in erroneous shadowing, yielding artificially jagged edges and temporally crawling shadows even using perspective optimization techniques. We introduce Sub-Pixel Shadow Maps (SPSM) for realtime shadow-mapping with sub-pixel precision. Our technique is based on the storage of a fixed-size partial representation of the scene geometry using conservative rasterization, combined with an original reconstruction of shadow edges. With only a small computational overhead our method avoids both perspective and projection aliasing. SPSM reconstructs shadow boundaries with thin details and high temporal consistency even using low resolution shadow maps. Our solution can also be used in conjunction with perspective optimization techniques where temporal and projection aliasing is hardly addressed. SPSM can particularly be used in demanding application such as games and real time rendering engines.

Comparison of Modern Omnidirectional Precise Shadowing Techniques Versus Ray Tracing

“…Ray tracing can achieve sub‐pixel precision naively by simply casting more secondary rays, or by beam [ORM07] or packet tracing [BEL*07]. Sub‐pixel shadow maps [LMSG14] try to address both perspective and projection alias of shadow maps by storing fixed‐sized partial representation of the scene geometry. Conservative rasterization is then used to generate fragments covering the entire triangle.…”

Revectorization‐Based Soft Shadow Mapping

“…In our proposal, rather than coupling the basis of PCSS with the anti‐aliasing provided by other techniques, such as [LMSG14, WHL15], we use the RBSM [MA16] because this technique is able to produce anti‐aliased hard and filtered hard shadows with only ∼1.2% of additional processing time when compared to shadow mapping [MTAA17]. Then, we replace the PCF step of PCSS by such a revectorization‐based filtering in the shadow map space to produce anti‐aliased real‐time soft shadows, even for low‐resolution shadow maps and small filter sizes.…”

Optimized Visibility Functions for Revectorization-Based Shadow Mapping

“…Silhouette Recovery: Some techniques aim to compute accurate shadow silhouettes either by proposing a hybrid approach [Hertel et al 2009] or embedding additional geometric data into the shadow map representation [Lecocq et al 2014]. In both cases, accurate shadows are generated at the cost of increased processing time and high memory consumption.…”

“…Also, an integration and evaluation of RBSM in the existing game engines would help on the popularization of the technique. Finally, a hybrid approach between RBSM and related work (such as [Lecocq et al 2014]) could be proposed to improve the accuracy of the revectorization.…”