Selective rendering

Abstract

The computational requirements of a full physically-based global illumination solution are significant, currently precluding its solution on even a powerful modern PC in reasonable let alone real time.A key factor to consider if we are ever to achieve so-called "Realism in Real-Time", is that we are computing images for humans to look at. Although the human visual system is very good, it is by no means perfect. By understanding what the human does, or perhaps more importantly, does not see, enables us to save significant computation effort without any loss of perceptual quality of the resultant image. This paper describes the novel techniques of selective rendering which allow us to direct computational resources to those areas of high perceptual importance while avoiding computing any detail which will not be perceived by the viewer. Such selective rendering methods offer us the real possibility of achieving high fidelity graphics of complex scenes at interactive rates.

DHR+S: distributed hybrid rendering with realistic real-time shadows for interactive thin client metaverse and game applications

“…The masked pixels are ray-traced as they are deemed to potentially be in some shadow penumbra, while the unmasked pixels marked in the shadow map are deemed to be in the umbra region. We are currently working on incorporating this a hybrid rendering effect that applies selective rendering [5] to avoid tracing shadow rays for every pixel.…”