Abstract While existing video and image quality datasets have extensively studied natural videos and traditional distortions, the perception of synthetic content and modern rendering artifacts remains underexplored. We present a novel video quality dataset focused on distortions introduced by advanced rendering techniques, including neural supersampling, novel‐view synthesis, path tracing, neural denoising, frame interpolation, and variable rate shading. Our evaluations show that existing full‐reference quality metrics perform sub‐optimally on these distortions, with a maximum Pearson correlation of 0.78. Additionally, we find that the feature space of pre‐trained 3D CNNs aligns strongly with human perception of visual quality. We propose CGVQM, a full‐reference video quality metric that significantly outperforms existing metrics while generating both per‐pixel error maps and global quality scores. Our dataset and metric implementation is available at https://github.com/IntelLabs/CGVQM .

Real-time rendering has been embracing ever-demanding effects, such as ray tracing. However, rendering such effects in high resolution and high frame rate remains challenging. Frame extrapolation methods, which do not introduce additional latency as opposed to frame interpolation methods such as DLSS 3 and FSR 3, boost the frame rate by generating future frames based on previous frames. However, it is a more challenging task because of the lack of information in the disocclusion regions and complex future motions, and recent methods also have a high engine integration cost due to requiring G-buffers as input. We propose a G-buffer free frame extrapolation method, GFFE, with a novel heuristic framework and an efficient neural network, to plausibly generate new frames in real time without introducing additional latency. We analyze the motion of dynamic fragments and different types of disocclusions, and design the corresponding modules of the extrapolation block to handle them. After that, a light-weight shading correction network is used to correct shading and improve overall quality. GFFE achieves comparable or better results than previous interpolation and G-buffer dependent extrapolation methods, with more efficient performance and easier integration.

We introduce ExtraSS, a novel framework that combines spatial super sampling and frame extrapolation to enhance real-time rendering performance. By integrating these techniques, our approach achieves a balance between performance and quality, generating temporally stable and high-quality, high-resolution results. Leveraging lightweight modules on warping and the ExtraSSNet for refinement, we exploit spatial-temporal information, improve rendering sharpness, handle moving shadings accurately, and generate temporally stable results. Computational costs are significantly reduced compared to traditional rendering methods, enabling higher frame rates and alias-free high resolution results. Evaluation using Unreal Engine demonstrates the benefits of our framework over conventional individual spatial or temporal super sampling methods, delivering improved rendering speed and visual quality. With its ability to generate temporally stable high-quality results, our framework creates new possibilities for real-time rendering applications, advancing the boundaries of performance and photo-realistic rendering in various domains.

Physically based rendering of complex scenes can be prohibitively costly with a potentially unbounded and uneven distribution of complexity across the rendered image. The goal of an ideal level of detail (LoD) method is to make rendering costs independent of the 3D scene complexity, while preserving the appearance of the scene. However, current prefiltering LoD methods are limited in the appearances they can support due to their reliance of approximate models and other heuristics. We propose the first comprehensive multi-scale LoD framework for prefiltering 3D environments with complex geometry and materials (e.g., the Disney BRDF), while maintaining the appearance with respect to the ray-traced reference. Using a multi-scale hierarchy of the scene, we perform a data-driven prefiltering step to obtain an appearance phase function and directional coverage mask at each scale. At the heart of our approach is a novel neural representation that encodes this information into a compact latent form that is easy to decode inside a physically based renderer. Once a scene is baked out, our method requires no original geometry, materials, or textures at render time. We demonstrate that our approach compares favorably to state-of-the-art prefiltering methods and achieves considerable savings in memory for complex scenes.

We introduce a practical general-purpose neural appearance filtering pipeline for physically-based rendering. We tackle the previously difficult challenge of aggregating visibility across many levels of detail from local information only, without relying on learning visibility for the entire scene. The high adaptivity of neural representations allows us to retain geometric correlations along rays and thus avoid light leaks. Common approaches to prefiltering decompose the appearance of a scene into volumetric representations with physically-motivated parameters, where the inflexibility of the fitted models limits rendering accuracy. We avoid assumptions on particular types of geometry or materials, bypassing any special-case decompositions. Instead, we directly learn a compressed representation of the intra-voxel light transport. For such high-dimensional functions, neural networks have proven to be useful representations. To satisfy the opposing constraints of prefiltered appearance and correlation-preserving point-to-point visibility, we use two small independent networks on a sparse multi-level voxel grid. Each network requires 10--20 minutes of training to learn the appearance of an asset across levels of detail. Our method achieves 70--95% compression ratios and around 25% of quality improvements over previous work. We reach interactive to real-time framerates, depending on the level of detail.

Abstract The recent research explosion around implicit neural representations, such as NeRF, shows that there is immense potential for implicitly storing high‐quality scene and lighting information in compact neural networks. However, one major limitation preventing the use of NeRF in real‐time rendering applications is the prohibitive computational cost of excessive network evaluations along each view ray, requiring dozens of petaFLOPS. In this work, we bring compact neural representations closer to practical rendering of synthetic content in real‐time applications, such as games and virtual reality. We show that the number of samples required for each view ray can be significantly reduced when samples are placed around surfaces in the scene without compromising image quality. To this end, we propose a depth oracle network that predicts ray sample locations for each view ray with a single network evaluation. We show that using a classification network around logarithmically discretized and spherically warped depth values is essential to encode surface locations rather than directly estimating depth. The combination of these techniques leads to DONeRF, our compact dual network design with a depth oracle network as its first step and a locally sampled shading network for ray accumulation. With DONeRF, we reduce the inference costs by up to 48× compared to NeRF when conditioning on available ground truth depth information. Compared to concurrent acceleration methods for raymarching‐based neural representations, DONeRF does not require additional memory for explicit caching or acceleration structures, and can render interactively (20 frames per second) on a single GPU.

FovVideoVDP is a video difference metric that models the spatial, temporal, and peripheral aspects of perception. While many other metrics are available, our work provides the first practical treatment of these three central aspects of vision simultaneously. The complex interplay between spatial and temporal sensitivity across retinal locations is especially important for displays that cover a large field-of-view, such as Virtual and Augmented Reality displays, and associated methods, such as foveated rendering. Our metric is derived from psychophysical studies of the early visual system, which model spatio-temporal contrast sensitivity, cortical magnification and contrast masking. It accounts for physical specification of the display (luminance, size, resolution) and viewing distance. To validate the metric, we collected a novel foveated rendering dataset which captures quality degradation due to sampling and reconstruction. To demonstrate our algorithm's generality, we test it on 3 independent foveated video datasets, and on a large image quality dataset, achieving the best performance across all datasets when compared to the state-of-the-art.

We introduce a novel transmittance model to improve the volumetric representation of 3D scenes. The model can represent opaque surfaces in the volumetric light transport framework. Volumetric representations are useful for complex scenes, and become increasingly popular for level of detail and scene reconstruction. The traditional exponential transmittance model found in volumetric light transport cannot capture correlations in visibility across volume elements. When representing opaque surfaces as volumetric density, this leads to both bloating of silhouettes and light leaking artifacts. By introducing a parametric non-exponential transmittance model, we are able to approximate these correlation effects and significantly improve the accuracy of volumetric appearance representation of opaque scenes. Our parametric transmittance model can represent a continuum between the linear transmittance that opaque surfaces exhibit and the traditional exponential transmittance encountered in participating media and unstructured geometries. This covers a large part of the spectrum of geometric structures encountered in complex scenes. In order to handle the spatially varying transmittance correlation effects, we further extend the theory of non-exponential participating media to a heterogeneous transmittance model. Our model is compact in storage and computationally efficient both for evaluation and for reverse-mode gradient computation. Applying our model to optimization algorithms yields significant improvements in volumetric scene appearance quality. We further show improvements for relevant applications, such as scene appearance prefiltering, image-based scene reconstruction using differentiable rendering, neural representations, and compare it to a conventional exponential model.

Due to higher resolutions and refresh rates, as well as more photorealistic effects, real-time rendering has become increasingly challenging for video games and emerging virtual reality headsets. To meet this demand, modern graphics hardware and game engines often reduce the computational cost by rendering at a lower resolution and then upsampling to the native resolution. Following the recent advances in image and video superresolution in computer vision, we propose a machine learning approach that is specifically tailored for high-quality upsampling of rendered content in real-time applications. The main insight of our work is that in rendered content, the image pixels are point-sampled, but precise temporal dynamics are available. Our method combines this specific information that is typically available in modern renderers (i.e., depth and dense motion vectors) with a novel temporal network design that takes into account such specifics and is aimed at maximizing video quality while delivering real-time performance. By training on a large synthetic dataset rendered from multiple 3D scenes with recorded camera motion, we demonstrate high fidelity and temporally stable results in real-time, even in the highly challenging 4 × 4 upsampling scenario, significantly outperforming existing superresolution and temporal antialiasing work.

Modern computer vision algorithms have brought significant advancement to 3D geometry reconstruction. However, illumination and material reconstruction remain less studied, with current approaches assuming very simplified models for materials and illumination. We introduce Inverse Path Tracing, a novel approach to jointly estimate the material properties of objects and light sources in indoor scenes by using an invertible light transport simulation. We assume a coarse geometry scan, along with corresponding images and camera poses. The key contribution of this work is an accurate and simultaneous retrieval of light sources and physically based material properties (e.g., diffuse reflectance, specular reflectance, roughness, etc.) for the purpose of editing and re-rendering the scene under new conditions. To this end, we introduce a novel optimization method using a differentiable Monte Carlo renderer that computes derivatives with respect to the estimated unknown illumination and material properties. This enables joint optimization for physically correct light transport and material models using a tailored stochastic gradient descent.

Shading filtering proposed by Kaplanyan et al. [2016] is a simple solution for specular aliasing. It filters a distribution of microfacet normals in the domain of microfacet slopes by estimating the filtering kernel using derivatives of a halfway vector between incident and outdoing directions. However, for real-time rendering, this approach can produce noticeable artifacts because of an estimation error of derivatives. For forward rendering, this estimation error is increased significantly at grazing angles and near edges. The present work improves the quality of the original technique, while decreasing the complexity of the code at the same time. To reduce the error, we introduce a more efficient kernel bandwidth that takes the angle of the halfvector into account. In addition, we optimize the calculation of an isotropic filter kernel used for deferred rendering by applying the proposed kernel bandwidth. As our implementation is simpler than the original method, it is easier to integrate in time-sensitive applications, such as game engines, while at the same time improving the filtering quality.

In order to provide an immersive visual experience, modern displays require head mounting, high image resolution, low latency, as well as high refresh rate. This poses a challenging computational problem. On the other hand, the human visual system can consume only a tiny fraction of this video stream due to the drastic acuity loss in the peripheral vision. Foveated rendering and compression can save computations by reducing the image quality in the peripheral vision. However, this can cause noticeable artifacts in the periphery, or, if done conservatively, would provide only modest savings. In this work, we explore a novel foveated reconstruction method that employs the recent advances in generative adversarial neural networks. We reconstruct a plausible peripheral video from a small fraction of pixels provided every frame. The reconstruction is done by finding the closest matching video to this sparse input stream of pixels on the learned manifold of natural videos. Our method is more efficient than the state-of-the-art foveated rendering, while providing the visual experience with no noticeable quality degradation. We conducted a user study to validate our reconstruction method and compare it against existing foveated rendering and video compression techniques. Our method is fast enough to drive gaze-contingent head-mounted displays in real time on modern hardware. We plan to publish the trained network to establish a new quality bar for foveated rendering and compression as well as encourage follow-up research.

Addressing vergence-accommodation conflict in head-mounted displays (HMDs) requires resolving two interrelated problems. First, the hardware must support viewing sharp imagery over the full accommodation range of the user. Second, HMDs should accurately reproduce retinal defocus blur to correctly drive accommodation. A multitude of accommodation-supporting HMDs have been proposed, with three architectures receiving particular attention: varifocal, multifocal, and light field displays. These designs all extend depth of focus, but rely on computationally expensive rendering and optimization algorithms to reproduce accurate defocus blur (often limiting content complexity and interactive applications). To date, no unified framework has been proposed to support driving these emerging HMDs using commodity content. In this paper, we introduce DeepFocus , a generic, end-to-end convolutional neural network designed to efficiently solve the full range of computational tasks for accommodation-supporting HMDs. This network is demonstrated to accurately synthesize defocus blur, focal stacks, multilayer decompositions, and multiview imagery using only commonly available RGB-D images, enabling real-time, near-correct depictions of retinal blur with a broad set of accommodation-supporting HMDs.

We describe a machine learning technique for reconstructing image sequences rendered using Monte Carlo methods. Our primary focus is on reconstruction of global illumination with extremely low sampling budgets at interactive rates. Motivated by recent advances in image restoration with deep convolutional networks, we propose a variant of these networks better suited to the class of noise present in Monte Carlo rendering. We allow for much larger pixel neighborhoods to be taken into account, while also improving execution speed by an order of magnitude. Our primary contribution is the addition of recurrent connections to the network in order to drastically improve temporal stability for sequences of sparsely sampled input images. Our method also has the desirable property of automatically modeling relationships based on auxiliary per-pixel input channels, such as depth and normals. We show significantly higher quality results compared to existing methods that run at comparable speeds, and furthermore argue a clear path for making our method run at realtime rates in the near future.

Rendering algorithms using Markov chain Monte Carlo (MCMC) currently build upon two different state spaces. One of them is the path space, where the algorithms operate on the vertices of actual transport paths. The other state space is the primary sample space, where the algorithms operate on sequences of numbers used for generating transport paths. While the two state spaces are related by the sampling procedure of transport paths, all existing MCMC rendering algorithms are designed to work within only one of the state spaces. We propose a first framework which provides a comprehensive connection between the path space and the primary sample space. Using this framework, we can use mutation strategies designed for one space with mutation strategies in the respective other space. As a practical example, we take a combination of manifold exploration and multiplexed Metropolis light transport using our framework. Our results show that the simultaneous use of the two state spaces improves the robustness of MCMC rendering. By combining efficient local exploration in the path space with global jumps in primary sample space, our method achieves more uniform convergence as compared to using only one space.

We present a stable shading method and a procedural shading model that enables real-time rendering of sub-pixel glints and anisotropic microdetails resulting from irregular microscopic surface structure to simulate a rich spectrum of appearances ranging from sparkling to brushed materials. We introduce a biscale Normal Distribution Function (NDF) for microdetails to provide a convenient artistic control over both the global appearance as well as over the appearance of the individual microdetail shapes, while efficiently generating procedural details. Our stable rendering approach simulates a hierarchy of scales and accurately estimates pixel footprint at multiple levels of detail to achieve good temporal stability and antialiasing, making it feasible for real-time rendering applications.

Foveated rendering synthesizes images with progressively less detail outside the eye fixation region, potentially unlocking significant speedups for wide field-of-view displays, such as head mounted displays, where target framerate and resolution is increasing faster than the performance of traditional real-time renderers. To study and improve potential gains, we designed a foveated rendering user study to evaluate the perceptual abilities of human peripheral vision when viewing today's displays. We determined that filtering peripheral regions reduces contrast, inducing a sense of tunnel vision. When applying a postprocess contrast enhancement, subjects tolerated up to 2× larger blur radius before detecting differences from a non-foveated ground truth. After verifying these insights on both desktop and head mounted displays augmented with high-speed gaze-tracking, we designed a perceptual target image to strive for when engineering a production foveated renderer. Given our perceptual target, we designed a practical foveated rendering system that reduces number of shades by up to 70% and allows coarsened shading up to 30° closer to the fovea than Guenter et al. [2012] without introducing perceivable aliasing or blur. We filter both pre- and post-shading to address aliasing from undersampling in the periphery, introduce a novel multiresolution- and saccade-aware temporal antialising algorithm, and use contrast enhancement to help recover peripheral details that are resolvable by our eye but degraded by filtering. We validate our system by performing another user study. Frequency analysis shows our system closely matches our perceptual target. Measurements of temporal stability show we obtain quality similar to temporally filtered non-foveated renderings.

Abstract In this paper, we present improvements to half vector space light transport (HSLT) [ KHD14 ], which make this approach more practical, robust for difficult input geometry, and faster. Our first contribution is the computation of half vector space ray differentials in a different domain than the original work. This enables a more uniform stratification over the image plane during Markov chain exploration. Furthermore, we introduce a new multi chain perturbation in half vector space, which, if combined appropriately with half vector perturbation, makes the mutation strategy both more robust to geometric configurations with fine displacements and faster due to reduced number of ray casts. We provide and analyze the results of improved HSLT and discuss possible applications of our new half vector ray differentials.

Global illumination algorithms using Markov chain Monte Carlo (MCMC) sampling are well-known for their efficiency in scenes with complex light transport. Samples in such algorithms are generated as a history of Markov chain states so that they are distributed according to the contributions to the image. The whole process is done based only on the information of the path contributions and user-defined transition probabilities from one state to the others. In light transport simulation, however, there is more information that can be used to improve the efficiency of path sampling. A notable example is multiple importance sampling (MIS) in bidirectional path tracing, which utilizes the probability densities of constructing a given path with different estimators. While MIS is a powerful ordinary Monte Carlo method, how to incorporate such additional information into MCMC sampling has been an open problem. We introduce a novel MCMC sampling framework, primary space serial tempering , which fuses the ideas of MCMC sampling and MIS for the first time. The key idea is to explore not only the sample space using a Markov chain, but also different estimators to generate samples by utilizing the information already available for MIS. Based on this framework, we also develop a novel rendering algorithm, multiplexed Metropolis light transport , which automatically and adaptively constructs paths with appropriate techniques as predicted by MIS. The final algorithm is very easy to implement, yet in many cases shows comparable (or even better) performance than significantly more complex MCMC rendering algorithms.

The path integral formulation of light transport is the basis for (Markov chain) Monte Carlo global illumination methods. In this paper we present half vector space light transport (HSLT) , a novel approach to sampling and integrating light transport paths on surfaces. The key is a partitioning of the path space into subspaces in which a path is represented by its start and end point constraints and a sequence of generalized half vectors. We show that this representation has several benefits. It enables importance sampling of all interactions along paths in between two endpoints. Based on this, we propose a new mutation strategy, to be used with Markov chain Monte Carlo methods such as Metropolis light transport (MLT), which is well-suited for all types of surface transport paths (diffuse/glossy/specular interaction). One important characteristic of our approach is that the Fourier-domain properties of the path integral can be easily estimated. These can be used to achieve optimal correlation of the samples due to well-chosen mutation step sizes, leading to more efficient exploration of light transport features. We also propose a novel approach to control stratification in MLT with our mutation strategy.

Abstract We propose a simple yet powerful regularization framework for robust light transport simulation. It builds on top of existing unbiased methods and resorts to a consistent estimation using regularization only for paths which cannot be sampled in an unbiased way. To introduce as little bias as possible, we selectively regularize individual interactions along paths, and also derive the regularization consistency conditions. Our approach is compatible with the majority of unbiased methods, e.g. (bidirectional) path tracing and Metropolis light transport (MLT), and only a simple modification is required to adapt existing renderers. We compare to recent unbiased and consistent methods and show examples of scenes with difficult light paths, where regularization is required to account for all illumination features. When coupled with MLT we are able to sample all phenomena, like recent consistent methods, while achieving superior convergence.

This article introduces a novellocally adaptiveprogressive photon mapping technique which optimally balances noise and bias in rendered images to minimize the overall error. It is the result of an analysis of the radiance estimation in progressive photon mapping. As a first step, we establish a connection to the field of recursive estimation and regression in statistics and derive the optimal estimation parameters for the asymptotic convergence of existing approaches. Next, we show how to reformulate photon mapping as a spatial regression in the measurement equation of light transport. This reformulation allows us to derive a novel data-driven bandwidth selection technique for estimating a pixel's measurement. The proposed technique possesses attractive convergence properties with finite numbers of samples, which is important for progressive rendering, and it also provides better results for quasi-converged images. Our results show the practical benefits of using our adaptive method.

Industry-quality content creation relies on tools for lighting artists to quickly prototype, iterate, and refine final renders. As industry-leading studios quickly adopt physically-based rendering (PBR) across their art generation pipelines, many existing tools have become unsuitable as they address only simple effects without considering underlying PBR concepts and constraints. We present a novel light transport manipulation technique that operates directly on path-space solutions of the rendering equation. We expose intuitive direct and indirect manipulation approaches to edit complex effects such as (multi-refracted) caustics, diffuse and glossy indirect bounces, and direct/indirect shadows. With our sketch- and object-space selection, all built atop a parameterized regular expression engine, artists can search and isolate shading effects to inspect and edit. We classify and filter paths on the fly and visualize the selected transport phenomena. We survey artists who used our tool to manipulate complex phenomena on both static and animated scenes.

Abstract This paper presents novel interactive visualization techniques for inspecting the global light transport in virtual scenes. First, we propose a simple extension to photon mapping to gather required lighting information. We then introduce a set of five light inspection tools which process this data to provide further insights. Corresponding visualizations help the user to comprehend how light travels within a scene, how the lighting affects the appearance of a surface, and how objects cause lighting effects such as caustics. We implemented all tools for direct usage in real production environments. Rendering is based on progressive photon mapping, providing interactivity and immediate visual feedback. We conducted a user study to evaluate all techniques in various application scenarios and hence discuss their individual strengths and weaknesses. Moreover, we present feedback from domain experts.

This paper introduces a new scalable technique for approximating indirect illumination in fully dynamic scenes for real-time applications, such as video games. We use lattices and spherical harmonics to represent the spatial and angular distribution of light in the scene. Our technique does not require any precomputation and handles large scenes with nested lattices. It is primarily targeted at rendering single-bounce indirect illumination with occlusion, but can be extended to handle multiple bounces and participating media. We demonstrate that our method produces plausible results even when running on current game console hardware with a budget of only a few milliseconds for performing all computation steps for indirect lighting. We evaluate our technique and show it in combination with a variety of popular real-time rendering techniques.