Figure 1: Decoupled deferred shading enables efficient shading reuse for stochastic rasterization. These images show depth of field rendering with 4× visibility supersampling (left), a visualization of the shading rate (sspp-shading samples per pixel, center), and the same shading as seen from a pin-hole camera (right). Our adaptive scheme reduces the shading frequency of defocused regions. AbstractIn this paper we present decoupled deferred shading: a rendering technique based on a new data structure called compact geometry buffer, which stores shading samples independently from the visibility. This enables caching and efficient reuse of shading computation, e.g. for stochastic rasterization techniques. In contrast to previous methods, our decoupled shading can be efficiently implemented on current graphics hardware. We describe two variants which differ in the way the shading samples are cached: the first maintains a single cache for the entire image in global memory, while the second pursues a tile-based approach leveraging local memory of the GPU's multiprocessors. We demonstrate the application of decoupled deferred shading to speed up the rendering in applications with stochastic supersampling, depth of field, and motion blur.

Neural networks are often quantized to use reduced-precision arithmetic, as it greatly improves their storage and computational costs. This approach is commonly used in image classification and natural language processing applications. However, using a quantized network for the reconstruction of HDR images can lead to a significant loss in image quality. In this paper, we introduce QW-Net , a neural network for image reconstruction, in which close to 95% of the computations can be implemented with 4-bit integers. This is achieved using a combination of two U-shaped networks that are specialized for different tasks, a feature extraction network based on the U-Net architecture, coupled to a filtering network that reconstructs the output image. The feature extraction network has more computational complexity but is more resilient to quantization errors. The filtering network, on the other hand, has significantly fewer computations but requires higher precision. Our network recurrently warps and accumulates previous frames using motion vectors, producing temporally stable results with significantly better quality than TAA, a widely used technique in current games.

FPS23 FPS 26 FPS Figure 1: Our method renders surface and volume caustics using approximate beam tracing. These results demonstrate two-sided refractions, inhomogeneous participating media as well as multi-bounce light-surface interactions rendered at real-time frame rates. AbstractCaustics are detailed patterns of light reflected or refracted on specular surfaces into participating media or onto surfaces. In this paper we present a novel adaptive and scalable algorithm for rendering surface and volume caustics in single-scattering participating media at real-time frame rates. Motivated by both caustic mapping and triangle-based volumetric methods, our technique captures the specular surfaces in light-space, but traces beams of light instead of single photons. The beams are adaptively generated from a grid projected from the light source onto the scene's surfaces, which is iteratively refined according to discontinuities in the geometry and photon distribution. This allows us to reconstruct sharp volume caustic patterns while reducing sampling resolution and fill-rate at defocused regions. We demonstrate our technique combined with approximate ray tracing techniques to render surfaces with twosided refractions as well as multiple caustic light bounces.

Recent advances in ray tracing hardware bring real-time path tracing into reach, and ray traced soft shadows, glossy reflections, and diffuse global illumination are now common features in games. Nonetheless, ray budgets are still limited. This results in undersampling, which manifests as aliasing and noise. Prior work addresses these issues separately. While temporal supersampling methods based on neural networks have gained a wide use in modern games due to their better robustness, neural denoising remains challenging because of its higher computational cost. We introduce a novel neural network architecture for real-time rendering that combines supersampling and denoising, thus lowering the cost compared to two separate networks. This is achieved by sharing a single low-precision feature extractor with multiple higher-precision filter stages. To reduce cost further, our network takes low-resolution inputs and reconstructs a high-resolution denoised supersampled output. Our technique produces temporally stable high-fidelity results that significantly outperform state-of-the-art real-time statistical or analytical denoisers combined with TAA or neural upsampling to the target resolution.

In this paper we present a fractional parametric splitting scheme for Reyes-style adaptive tessellation. Our parallel algorithm generates crack-free tessellation from a parametric surface, which is also free of sudden temporal changes under animation. Continuous level of detail is not addressed by existing Reyes-style methods, since these aim to produce subpixel-sized micropolygons, where topology changes are no longer noticeable. Using our method, rendering pipelines that use larger triangles, thus sensitive to geometric popping, may also benefit from the quality of the split-dice tessellation stages of Reyes. We demonstrate results on a real-time GPU implementation, going beyond the limited quality and resolution of the hardware tessellation unit. In contrast to previous split-dice methods, our split stage is compatible with the fractional hardware tessellation scheme that has been designed for continuous level of detail.