Prefiltered single scattering

Abstract

2.2ms 30fpsFigure 1: Our method needs only 2.2ms for single scattering using a 1024 2 shadow map and a 1920 × 1080 full-HD screen resolution. It has basically a constant cost per screen pixel, achieving up to a 20× speedup and 6× less memory consumption for similar quality compared to state-of-the-art methods that struggle with the complexity and details of the scene [Chen et al. 2011]. (Total render time is around 33 ms -most time is spent on ambient occlusion, alpha matting, and disabled backface culling to avoid geometry errors in the scene). AbstractVolumetric light scattering is a complex phenomenon that is difficult to simulate in real time as light can be scattered towards the camera from everywhere in space. By assuming a single-scattering model, we can transform the usually-employed ray-marching into an efficient ray-independent texture filtering process. Our algorithm builds upon a rectified shadow map as input and we propose an efficient rectification scheme, which could be used by other approaches as well. The resulting scattering method is very fast and almost independent of the screen resolution, but it still produces near-reference results. These properties make it a good candidate for performance-critical applications, such as games.

Efficient Algorithms for Real-Time GPU Volumetric Cloud Rendering with Enhanced Geometry

“…Our method also improves on the overall realism in the model of Elek et al [20], while maintaining the same performance as the conservative ray-marching version of their implementation when executed at 1920 × 1080 pixels. Despite our efforts, our method lacks the precision of the excellent work by Klehm et al [23] and Peters et al [24] with respect to scattering and shadow maps.…”

“…Volumetric rendering of clouds still requires a heavy workload for most modern GPUs, but the realism is very accurate and may become standard in the coming years. Later contributions to this approach were made by Yusov [19], Elek et al [20], Dobashi et al [21], and Goswami-Neyret [22], while further contributions include the latest cutting-edge findings in light scattering made by Klehm et al [23] and Peters et al [24], with very high performance at 1920 × 1080 pixels. We contributed to the state-of-the-art by improving the performance of existing algorithms and developing new efficient algorithms and methods to reduce the physics workload of many slow hyper-realistic implementations, without loss of quality.…”

Expressive Single Scattering for Light Shaft Stylization

“…Hence, we focus on existing real-time solutions for single scattering. There are several options for efficient computation of single scattering; min-max mipmaps [1], voxelized shadow volumes [21], shadow volumes based on shadow maps [22], or prefiltered single scattering [2]. While approaches for multiple scattering exist [23], their precision is still relatively low due to the use of a coarse grid, which is why we concentrate on single scattering only.…”

“…Note that we skip attenuation from the light source to the sampling point x t , as this results in complete attenuation for a directional light source located at infinity. A common approximation is to factor out visibility [1], [2], [21], [30], which we also identify as a useful parameter for stylization purposes. The equation then becomes:…”

“…Simulating single scattering in homogeneous media can be achieved in real time [1], [2]. As in nature, light shafts depend on the scene layout and not on any specific scene element, which makes stylization difficult as the appearance cannot be modified directly.…”