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.

Decoupled sampling techniques such as coarse pixel shading can lower the shading rate while resolving visibility at the full resolution, thereby preserving details along geometric edges. However, while these techniques can significantly reduce shading costs, they also reduce shading quality. In this paper we extend coarse pixel shading with a temporal supersampling scheme that notably improves image quality. We derive multiple shading and visibility samples by jittering each frame using a novel sequence that produces a suitable distribution of samples in both the shading and visibility domain, and temporally resolve these samples to enhance the shading resolution and reduce aliasing. We demonstrate a substantial reduction in shading cost compared to checkerboard rendering, which is another temporal supersampling technique widely used in games today.

Abstract In this paper, we extend the concept of pre‐filtered shadow mapping to stochastic rasterization, enabling real‐time rendering of soft shadows from planar area lights. Most existing soft shadow mapping methods lose important visibility information by relying on pinhole renderings from an area light source, providing plausible results only for small light sources. Since we sample the entire 4D shadow light field stochastically, we are able to closely approximate shadows of large area lights as well. In order to efficiently reconstruct smooth shadows from this sparse data, we exploit the analogy of soft shadow computation to rendering defocus blur, and introduce a multiplane pre‐filtering algorithm. We demonstrate how existing pre‐filterable approximations of the visibility function, such as variance shadow mapping, can be extended to four dimensions within our framework.

Abstract 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.

In 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.

Caustics 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 two-sided refractions as well as multiple caustic light bounces.

This paper presents a fast parallel method to solve the radiative transport equation in inhomogeneous participating media. We apply a novel approximation scheme to find a good initial guess for both the direct and scattered components. Then, the initial approximation is used to bootstrap an iterative multiple scattering solver, i.e., we let the iteration concentrate just on the residual problem. This kind of bootstrapping makes the volumetric source approximation more uniform, thus it helps to reduce the discretization artifacts and improves the efficiency of the parallel implementation. The iterative refinement is executed on a face-centered cubic grid. The implementation is based on CUDA and runs on the GPU. For large volumes that do not fit into the GPU memory, we also consider the implementation on a GPU cluster, where the volume is decomposed to blocks according to the available GPU nodes. We show how the communication bottleneck can be avoided in the cluster implementation by not exchanging the boundary conditions in every iteration step. In addition to light photons, we also discuss the generalization of the method to γ-photons that are relevant in medical simulation.