Abstract Rendering 3D virtual scenarios has become a popular alternative for generating per-pixel-labeled image datasets, especially in fields like autonomous driving. The approach is valuable for training neural perception models, such as semantic segmentation models, particularly when data might be scarce, expensive, or difficult to collect. However, fundamental questions persist within the research community regarding the generation and processing of these synthetic images, particularly a better understanding of the key factors influencing the performance of deep learning models trained with such synthetic images. In response, we conducted a series of experiments to elucidate the impact that common aspects involved in the generation of rendered synthetic images may have on the performance of neural semantic segmentation tasks. Our study used a recent autonomous driving synthetic dataset as our main testbed, allowing us to investigate the effect of different approaches when modeling their geometric, material, and lighting details. We also studied the impact of rendering noise, typically produced by path-tracing algorithms, as well as the impact of using different color transformations and tonemapping algorithms.

Scientific visualizations using physically-based lighting models play a crucial role in enhancing both image quality and realism. In the domain of medical visualization, this trend has gained significant traction under the term cinematic rendering (CR). It enables the creation of 3D photorealistic reconstructions from medical data, offering great potential for aiding healthcare professionals in the analysis and study of volumetric datasets. However, the adoption of such advanced rendering for immersive virtual reality (VR) faces two main limitations related to their high computational demands. First, these techniques are frequently used to produce pre-recorded videos and offline content, thereby restricting interactivity to predefined volume appearance and lighting settings. Second, when deployed in head-tracked VR environments they can induce cyber-sickness symptoms due to the disturbing flicker caused by noisy Monte Carlo renderings. Consequently, the scope for meaningful interactive operations is constrained in this modality, in contrast with the versatile capabilities of classical direct volume rendering (DVR). In this work, we introduce an immersive 3D medical visualization system capable of producing photorealistic and fully interactive stereoscopic visualizations on head-mounted display (HMD) devices. Our approach extends previous linear regression denoising to enable real-time stereoscopic cinematic rendering within AR/VR settings. We demonstrate the capabilities of the resulting VR system, like its interactive rendering, appearance and transfer function editing.

Abstract Real-time play recognition and classification algorithms are crucial for automating video production and live broadcasts of sporting events. However, current methods relying on human pose estimation and deep neural networks introduce high latency on commodity hardware, limiting their usability in low-cost real-time applications. We present PlayNet, a novel approach to real-time handball play classification. Our method is based on Kalman embeddings, a new low-dimensional representation for game states that enables efficient operation on commodity hardware and customized camera layouts. Firstly, we leverage Kalman filtering to detect and track the main agents in the playing field, allowing us to represent them in a single normalized coordinate space. Secondly, we utilize a neural network trained in nonlinear dimensionality reduction through fuzzy topological data structure analysis. As a result, PlayNet achieves real-time play classification with under 55 ms of latency on commodity hardware, making it a promising addition to automated live broadcasting and game analysis pipelines.

Unbiased rendering algorithms such as path tracing produce accurate images given a huge number of samples, but in practice, the techniques often leave visually distracting artifacts (i.e., noise) in their rendered images due to a limited time budget. A favored approach for mitigating the noise problem is applying learning-based denoisers to unbiased but noisy rendered images and suppressing the noise while preserving image details. However, such denoising techniques typically introduce a systematic error, i.e., the denoising bias, which does not decline as rapidly when increasing the sample size, unlike the other type of error, i.e., variance. It can technically lead to slow numerical convergence of the denoising techniques. We propose a new combination framework built upon the James-Stein (JS) estimator, which merges a pair of unbiased and biased rendering images, e.g., a path-traced image and its denoised result. Unlike existing post-correction techniques for image denoising, our framework helps an input denoiser have lower errors than its unbiased input without relying on accurate estimation of per-pixel denoising errors. We demonstrate that our framework based on the well-established JS theories allows us to improve the error reduction rates of state-of-the-art learning-based denoisers more robustly than recent post-denoisers.

Direct volume rendering (DVR) using volumetric path tracing (VPT) is a scientific visualization technique that simulates light transport with objects' matter using physically-based lighting models. Monte Carlo (MC) path tracing is often used with surface models, yet its application for volumetric models is difficult due to the complexity of integrating MC light-paths in volumetric media with none or smooth material boundaries. Moreover, auxiliary geometry-buffers (G-buffers) produced for volumes are typically very noisy, failing to guide image denoisers relying on that information to preserve image details. This makes existing real-time denoisers, which take noise-free G-buffers as their input, less effective when denoising VPT images. We propose the necessary modifications to an image-based denoiser previously used when rendering surface models, and demonstrate effective denoising of VPT images. In particular, our denoising exploits temporal coherence between frames, without relying on noise-free G-buffers, which has been a common assumption of existing denoisers for surface-models. Our technique preserves high-frequency details through a weighted recursive least squares that handles heterogeneous noise for volumetric models. We show for various real data sets that our method improves the visual fidelity and temporal stability of VPT during classic DVR operations such as camera movements, modifications of the light sources, and editions to the volume transfer function.

Variable rate is a requirement for flexible and adaptable image and video compression. However, deep image compression methods (DIC) are optimized for a single fixed rate-distortion (R-D) tradeoff. While this can be addressed by training multiple models for different tradeoffs, the memory requirements increase proportionally to the number of models. Scaling the bottleneck representation of a shared autoencoder can provide variable rate compression with a single shared autoencoder. However, the R-D performance using this simple mechanism degrades in low bitrates, and also shrinks the effective range of bitrates. To address these limitations, we formulate the problem of variable R-D optimization for DIC, and propose modulated autoencoders (MAEs), where the representations of a shared autoencoder are adapted to the specific R-D tradeoff via a modulation network. Jointly training this modulated autoencoder and the modulation network provides an effective way to navigate the R-D operational curve. Our experiments show that the proposed method can achieve almost the same R-D performance of independent models with significantly fewer parameters.

Abstract We propose a novel pre‐filtering method that reduces the noise introduced by depth‐of‐field and motion blur effects in geometric buffers (G‐buffers) such as texture, normal and depth images. Our pre‐filtering uses world positions and their variances to effectively remove high‐frequency noise while carefully preserving high‐frequency edges in the G‐buffers. We design a new anisotropic filter based on a per‐pixel covariance matrix of world position samples. A general error estimator, Stein's unbiased risk estimator, is then applied to estimate the optimal trade‐off between the bias and variance of pre‐filtered results. We have demonstrated that our pre‐filtering improves the results of existing filtering methods numerically and visually for challenging scenes where depth‐of‐field and motion blurring introduce a significant amount of noise in the G‐buffers.

Perceptually lossless foveated rendering methods exploit human perception by selectively rendering at different quality levels based on eye gaze (at a lower computational cost) while still maintaining the user's perception of a full quality render. We consider three foveated rendering methods and propose practical rules of thumb for each method to achieve significant performance gains in real-time rendering frameworks. Additionally, we contribute a new metric for perceptual foveated rendering quality building on HDR-VDP2 that, unlike traditional metrics, considers the loss of fidelity in peripheral vision by lowering the contrast sensitivity of the model with visual eccentricity based on the Cortical Magnification Factor (CMF). The new metric is parameterized on user-test data generated in this study. Finally, we run our metric on a novel foveated rendering method for real-time immersive 360° content with motion parallax.

We address the problem of denoising Monte Carlo renderings by studying existing approaches and proposing a new algorithm that yields state-of-the-art performance on a wide range of scenes. We analyze existing approaches from a theoretical and empirical point of view, relating the strengths and limitations of their corresponding components with an emphasis on production requirements. The observations of our analysis instruct the design of our new filter that offers high-quality results and stable performance. A key observation of our analysis is that using auxiliary buffers (normal, albedo, etc.) to compute the regression weights greatly improves the robustness of zero-order models, but can be detrimental to first-order models. Consequently, our filter performs a first-order regression leveraging a rich set of auxiliary buffers only when fitting the data, and, unlike recent works, considers the pixel color alone when computing the regression weights. We further improve the quality of our output by using a collaborative denoising scheme. Lastly, we introduce a general mean squared error estimator, which can handle the collaborative nature of our filter and its nonlinear weights, to automatically set the bandwidth of our regression kernel.

Abstract We propose a new real‐time temporal filtering and antialiasing (AA) method for rasterization graphics pipelines. Our method is based on Pixel History Linear Models (PHLM), a new concept for modeling the history of pixel shading values over time using linear models. Based on PHLM, our method can predict per‐pixel variations of the shading function between consecutive frames. This combines temporal reprojection with per‐pixel shading predictions in order to provide temporally coherent shading, even in the presence of very noisy input images. Our method can address both spatial and temporal aliasing problems under a unique filtering framework that minimizes filtering error through a recursive least squares algorithm. We demonstrate our method working with a commercial deferred shading engine for rasterization and with our own OpenGL deferred shading renderer. We have implemented our method in GPU and it has shown significant reduction of temporal flicker in very challenging scenarios including foliage rendering, complex non‐linear camera motions, dynamic lighting, reflections, shadows and fine geometric details. Our approach, based on PHLM, avoids the creation of visible ghosting artifacts and it reduces the filtering overblur characteristic of temporal deflickering methods. At the same time, the results are comparable to state‐of‐the‐art real‐time filters in terms of temporal coherence.

Abstract This paper presents a time‐varying, multi‐layered biophysically‐based model of the optical properties of human skin, suitable for simulating appearance changes due to aging. We have identified the key aspects that cause such changes, both in terms of the structure of skin and its chromophore concentrations, and rely on the extensive medical and optical tissue literature for accurate data. Our model can be expressed in terms of biophysical parameters, optical parameters commonly used in graphics and rendering (such as spectral absorption and scattering coefficients), or more intuitively with higher‐level parameters such as age, gender, skin care or skin type. It can be used with any rendering algorithm that uses diffusion profiles, and it allows to automatically simulate different types of skin at different stages of aging, avoiding the need for artistic input or costly capture processes. While the presented skin model is inspired on tissue optics studies, we also provided a simplified version valid for non‐diagnostic applications.

We propose a new adaptive rendering algorithm that enhances the performance of Monte Carlo ray tracing by reducing the noise, i.e., variance, while preserving a variety of high-frequency edges in rendered images through a novel prediction based reconstruction. To achieve our goal, we iteratively build multiple, but sparse linear models. Each linear model has its prediction window, where the linear model predicts the unknown ground truth image that can be generated with an infinite number of samples. Our method recursively estimates prediction errors introduced by linear predictions performed with different prediction windows, and selects an optimal prediction window minimizing the error for each linear model. Since each linear model predicts multiple pixels within its optimal prediction interval, we can construct our linear models only at a sparse set of pixels in the image screen. Predicting multiple pixels with a single linear model poses technical challenges, related to deriving error analysis for regions rather than pixels, and has not been addressed in the field. We address these technical challenges, and our method with robust error analysis leads to a drastically reduced reconstruction time even with higher rendering quality, compared to state-of-the-art adaptive methods. We have demonstrated that our method outperforms previous methods numerically and visually with high performance ray tracing kernels such as OptiX and Embree.

Abstract Great advancements in commodity graphics hardware have favoured graphics processing unit (GPU)‐based volume rendering as the main adopted solution for interactive exploration of rectilinear scalar volumes on commodity platforms. Nevertheless, long data transfer times and GPU memory size limitations are often the main limiting factors, especially for massive, time‐varying or multi‐volume visualization, as well as for networked visualization on the emerging mobile devices. To address this issue, a variety of level‐of‐detail (LOD) data representations and compression techniques have been introduced. In order to improve capabilities and performance over the entire storage, distribution and rendering pipeline, the encoding/decoding process is typically highly asymmetric, and systems should ideally compress at data production time and decompress on demand at rendering time. Compression and LOD pre‐computation does not have to adhere to real‐time constraints and can be performed off‐line for high‐quality results. In contrast, adaptive real‐time rendering from compressed representations requires fast, transient and spatially independent decompression. In this report, we review the existing compressed GPU volume rendering approaches, covering sampling grid layouts, compact representation models, compression techniques, GPU rendering architectures and fast decoding techniques.

Abstract We present a novel multiresolution compression‐domain GPU volume rendering architecture designed for interactive local and networked exploration of rectilinear scalar volumes on commodity platforms. In our approach, the volume is decomposed into a multiresolution hierarchy of bricks. Each brick is further subdivided into smaller blocks, which are compactly described by sparse linear combinations of prototype blocks stored in an overcomplete dictionary. The dictionary is learned, using limited computational and memory resources, by applying the K‐SVD algorithm to a re‐weighted non‐uniformly sampled subset of the input volume, harnessing the recently introduced method of coresets. The result is a scalable high quality coding scheme, which allows very large volumes to be compressed off‐line and then decompressed on‐demand during real‐time GPU‐accelerated rendering. Volumetric information can be maintained in compressed format through all the rendering pipeline. In order to efficiently support high quality filtering and shading, a specialized real‐time renderer closely coordinates decompression with rendering, combining at each frame images produced by raycasting selectively decompressed portions of the current view‐ and transfer‐function‐dependent working set. The quality and performance of our approach is demonstrated on massive static and time‐varying datasets.

Large scale and structurally complex volume datasets from high-resolution 3D imaging devices or computational simulations pose a number of technical challenges for interactive visual analysis. In this paper, we present the first integration of a multiscale volume representation based on tensor approximation within a GPU-accelerated out-of-core multiresolution rendering framework. Specific contributions include (a) a hierarchical brick-tensor decomposition approach for pre-processing large volume data, (b) a GPU accelerated tensor reconstruction implementation exploiting CUDA capabilities, and (c) an effective tensor-specific quantization strategy for reducing data transfer bandwidth and out-of-core memory footprint. Our multiscale representation allows for the extraction, analysis and display of structural features at variable spatial scales, while adaptive level-of-detail rendering methods make it possible to interactively explore large datasets within a constrained memory footprint. The quality and performance of our prototype system is evaluated on large structurally complex datasets, including gigabyte-sized micro-tomographic volumes.

Abstract We present a GPU accelerated volume ray casting system interactively driving a multi‐user light field display. The display, driven by a single programmable GPU, is based on a specially arranged array of projectors and a holographic screen and provides full horizontal parallax. The characteristics of the display are exploited to develop a specialized volume rendering technique able to provide multiple freely moving naked‐eye viewers the illusion of seeing and manipulating virtual volumetric objects floating in the display workspace. In our approach, a GPU ray‐caster follows rays generated by a multiple‐center‐of‐projection technique while sampling pre‐filtered versions of the dataset at resolutions that match the varying spatial accuracy of the display. The method achieves interactive performance and provides rapid visual understanding of complex volumetric data sets even when using depth oblivious compositing techniques.