Robust Image Denoising Using a Virtual Flash Image for Monte Carlo Ray Tracing

We propose an efficient and robust image‐space denoising method for noisy images generated by Monte Carlo ray tracing methods. Our method is based on two new concepts: virtual flash images and homogeneous pixels. Inspired by recent developments in flash photography, virtual flash images emulate photographs taken with a flash, to capture various features of rendered images without taking additional samples. Using a virtual flash image as an edge‐stopping function, our method can preserve image features that were not captured well only by existing edge‐stopping functions such as normals and depth values. While denoising each pixel, we consider only homogeneous pixels—pixels that are statistically equivalent to each other. This makes it possible to define a stochastic error bound of our method, and this bound goes to zero as the number of ray samples goes to infinity, irrespective of denoising parameters. To highlight the benefits of our method, we apply our method to two Monte Carlo ray tracing methods, photon mapping and path tracing, with various input scenes. We demonstrate that using virtual flash images and homogeneous pixels with a standard denoising method outperforms state‐of‐the‐art image‐space denoising methods.     

Global Illumination using Photon Ray Splatting

We present a novel framework for efficiently computing the indirect illumination in diffuse and moderately glossy scenes using density estimation techniques. Many existing global illumination approaches either quickly compute an overly approximate solution or perform an orders of magnitude slower computation to obtain high-quality results for the indirect illumination. The proposed method improves photon density estimation and leads to significantly better visual quality in particular for complex geometry, while only slightly increasing the computation time. We perform direct splatting of photon rays, which allows us to use simpler search data structures. Since our density estimation is carried out in ray space rather than on surfaces, as in the commonly used photon mapping algorithm, the results are more robust against geometrically incurred sources of bias. This holds also in combination with final gathering where photon mapping often overestimates the illumination near concave geometric features. In addition, we show that our photon splatting technique can be extended to handle moderately glossy surfaces and can be combined with traditional irradiance caching for sparse sampling and filtering in image space.     

An Eyeglass Simulator Using Conoid Tracing

This paper proposes a method for displaying images at the fovea of the retina taking visual acuity into account. Previous research has shown that a point light source projected onto the retina forms an ellipse, which can be computed with wavefront tracing from each point in space. We propose a novel concept using conoid tracing, with which we can acquire defocusing information several times faster than that acquired by previous methods. We also show that conoid tracing is more robust and produces higher quality results. In conoid tracing the ray is regarded as a conoid, a thin cone‐like shape with varying elliptical cross‐section. The viewing ray from the retina is traced as a conoid and evaluated at each sample location. Using the sampled and pre‐computed data for the spatial distribution of blurring, we implemented an interactive eyeglass simulator. This paper demonstrates some visualization results utilizing the interactivity of the simulator, which an eyeglass lens design company uses to evaluate the design of complex progressive lenses.     

Compact,Fast and Robust Grids for Ray Tracing

The focus of research in acceleration structures for ray tracing recently shifted from render time to time to image, the sum of build time and render time, and also the memory footprint of acceleration structures now receives more attention. In this paper we revisit the grid acceleration structure in this setting. We present two efficient methods for representing and building a grid. The compact grid method consists of a static data structure for representing a grid with minimal memory requirements, more specifically exactly one index per grid cell and exactly one index per object reference, and an algorithm for building that data structure in linear time. The hashed grid method reduces memory requirements even further, by using perfect hashing based on row displacement compression. We show that these methods are more efficient in both time and space than traditional methods based on linked lists and dynamic arrays. We also present a more robust grid traversal algorithm. We show that, for applications where time to image or memory usage is important, such as interactive ray tracing and rendering large models, the grid acceleration structure is an attractive alternative.     

Interactive Global Photon Mapping

We present a photon mapping technique capable of computing high quality global illumination at interactive frame rates. By extending the concept of photon differentials to efficiently handle diffuse reflections, we generate footprints at all photon hit points. These enable illumination reconstruction by density estimation with variable kernel bandwidths without having to locate the k nearest photon hits first. Adapting an efficient BVH construction process for ray tracing acceleration, we build photon maps that enable the fast retrieval of all hits relevant to a shading point. We present a heuristic that automatically tunes the BVH build's termination criterion to the scene and illumination conditions. As all stages of the algorithm are highly parallelizable, we demonstrate an implementation using NVidia's CUDA manycore architecture running at interactive rates on a single GPU. Both light source and camera may be freely moved with global illumination fully recalculated in each frame.     

Direct Ray Tracing of Phong Tessellation

There are two major ways of calculating ray and parametric surface intersections in rendering. The first is through the use of tessellated triangles, and the second is to use parametric surfaces together with numerical methods such as Newton's method. Both methods are computationally expensive and complicated to implement. In this paper, we focus on Phong Tessellation and introduce a simple direct ray tracing method for Phong Tessellation. Our method enables rendering smooth surfaces in a computationally inexpensive yet robust way.     

Photon Differential Splatting for Rendering Caustics

We present a photon splatting technique which reduces noise and blur in the rendering of caustics. Blurring of illumination edges is an inherent problem in photon splatting, as each photon is unaware of its neighbours when being splatted. This means that the splat size is usually based on heuristics rather than knowledge of the local flux density. We use photon differentials to determine the size and shape of the splats such that we achieve adaptive anisotropic flux density estimation in photon splatting. As compared to previous work that uses photon differentials, we present the first method where no photons or beams or differentials need to be stored in a map. We also present improvements in the theory of photon differentials, which give more accurate results and a faster implementation. Our technique has good potential for GPU acceleration, and we limit the number of parameters requiring user adjustment to an overall smoothing parameter and the number of photons to be traced.     

Improving Memory Space Efficiency of Kd‐tree for Real‐time Ray Tracing

Compared with its competitors such as the bounding volume hierarchy, a drawback of the kd‐tree structure is that a large number of triangles are repeatedly duplicated during its construction, which often leads to inefficient, large and tall binary trees with high triangle redundancy. In this paper, we propose a space‐efficient kd‐tree representation where, unlike commonly used methods, an inner node is allowed to optionally store a reference to a triangle, so highly redundant triangles in a kd‐tree can be culled from the leaf nodes and moved to the inner nodes. To avoid the construction of ineffective kd‐trees entailing computational inefficiencies due to early, possibly unnecessary, ray‐triangle intersection calculations that now have to be performed in the inner nodes during the kd‐tree traversal, we present heuristic measures for determining when and how to choose triangles for inner nodes during kd‐tree construction. Based on these metrics, we describe how the new form of kd‐tree is constructed and stored compactly using a carefully designed data layout. Our experiments with several example scenes showed that our kd‐tree representation technique significantly reduced the memory requirements for storing the kd‐tree structure, while effectively suppressing the unavoidable frame‐rate degradation observed during ray tracing.     

Importance Sampling Techniques for Path Tracing in Participating Media

We introduce a set of robust importance sampling techniques which allow efficient calculation of direct and indirect lighting from arbitrary light sources in both homogeneous and heterogeneous media. We show how to distribute samples along a ray proportionally to the incoming radiance for point and area lights. In heterogeneous media, we decouple ray marching from light calculations by computing a representation of the transmittance function that can be quickly evaluated during sampling, at the cost of a small amount of bias. This representation also allows the calculation of another probability density function which can direct samples to regions most likely to scatter light. These techniques are orthogonal and can be combined via multiple importance sampling to further reduce variance. Our method has very modest per‐ray memory requirements and does not require any preprocessing, making it simple to integrate into production ray tracing based renderers.     

Accurate Computation of Single Scattering in Participating Media with Refractive Boundaries

Volume caustics are high‐frequency effects appearing in participating media with low opacity, when refractive interfaces are focusing the light rays. Refractions make them hard to compute, since screen locality does not correlate with spatial locality in the medium. In this paper, we give a new method for accurate computation of single scattering effects in a participating media enclosed by refractive interfaces. Our algorithm is based on the observation that although radiance along each camera ray is irregular, contributions from individual triangles are smooth. Our method gives more accurate results than existing methods, faster. It uses minimal information and requires no pre‐computation or additional data structures.     

Crack‐free rendering of dynamically tesselated B‐rep models

We propose a versatile pipeline to render B‐Rep models interactively, precisely and without rendering‐related artifacts such as cracks. Our rendering method is based on dynamic surface evaluation using both tesselation and ray‐casting, and direct GPU surface trimming. An initial rendering of the scene is performed using dynamic tesselation. The algorithm we propose reliably detects then fills up cracks in the rendered image. Crack detection works in image space, using depth information, while crack‐filling is either achieved in image space using a simple classification process, or performed in object space through selective ray‐casting. The crack filling method can be dynamically changed at runtime. Our image space crack filling approach has a limited runtime cost and enables high quality, real‐time navigation. Our higher quality, object space approach results in a rendering of similar quality than full‐scene ray‐casting, but is 2 to 6 times faster, can be used during navigation and provides accurate, reliable rendering. Integration of our work with existing tesselation‐based rendering engines is straightforward.     

Efficient Monte Carlo rendering with realistic lenses

In this paper we present a novel approach to simulate image formation for a wide range of real world lenses in the Monte Carlo ray tracing framework. Our approach sidesteps the overhead of tracing rays through a system of lenses and requires no tabulation. To this end we first improve the precision of polynomial optics to closely match ground‐truth ray tracing. Second, we show how the Jacobian of the optical system enables efficient importance sampling, which is crucial for difficult paths such as sampling the aperture which is hidden behind lenses on both sides. Our results show that this yields converged images significantly faster than previous methods and accurately renders complex lens systems with negligible overhead compared to simple models, e.g. the thin lens model. We demonstrate the practicality of our method by incorporating it into a bidirectional path tracing framework and show how it can provide information needed for sophisticated light transport algorithms.     

Polynomial Optics: A Construction Kit for Efficient Ray‐Tracing of Lens Systems

Simulation of light transport through lens systems plays an important role in graphics. While basic imaging properties can be conveniently derived from linear models (like ABCD matrices), these approximations fail to describe nonlinear effects and aberrations that arise in real optics. Such effects can be computed by proper ray tracing, for which, however, finding suitable sampling and filtering strategies is often not a trivial task. Inspired by aberration theory, which describes the deviation from the linear ray transfer in terms of wavefront distortions, we propose a ray‐space formulation for nonlinear effects. In particular, we approximate the analytical solution to the ray tracing problem by means of a Taylor expansion in the ray parameters. This representation enables a construction‐kit approach to complex optical systems in the spirit of matrix optics. It is also very simple to evaluate, which allows for efficient execution on CPU and GPU alike, including the computation of mixed derivatives of any order. We evaluate fidelity and performance of our polynomial model, and show applications in high‐quality offline rendering and at interactive frame rates.     

Robust linear and support vector regression

The robust Huber M-estimator, a differentiable cost function that is quadratic for small errors and linear otherwise, is modeled exactly, in the original primal space of the problem, by an easily solvable simple convex quadratic program for both linear and nonlinear support vector estimators. Previous models were significantly more complex or formulated in the dual space and most involved specialized numerical algorithms for solving the robust Huber linear estimator. Numerical test comparisons with these algorithms indicate the computational effectiveness of the new quadratic programming model for both linear and nonlinear support vector problems. Results are shown on problems with as many as 20000 data points, with considerably faster running times on larger problems     

Accelerating Volume Raycasting using Proxy Spheres

In this paper, we propose an efficient solution that addresses the performance problems of current single-pass GPU raycasting algorithms. Our paper provides more control over the rendering process by introducing tighter ray segments for raycasting, while at the same time avoiding the introduction of any new rendering artefacts. We achieve this by dynamically generating, on the GPU, a coarsely fitted proxy geometry, composed of spheres, for the active blocks. The spheres are then rasterised into two z-buffers by a single rendering pass. The resulting two z-buffers are used as the first-hit and last-hit points for the subsequent raycaster. With this approach, only the valid ray segments between the two z-buffers need to be sampled during raycasting. This also provides more coherent parallelism on the GPU due to more consistent ray length and avoidance of the overheads and dynamic branching of performing checks on a per-sample basis during the raycasting pass.
Our technique is ideal for dynamic data exploration in which both the transfer function and view parameters need to be changed frequently at runtime. The rendering results of our algorithm are identical to the general cube-based proxy geometry algorithm, but the performance can be up to 15.7 times faster. Furthermore, the approach can be adopted by any existing raycasting system in a straightforward way.     

Ray Casting Algebraic Surfaces using the Frustum Form

We propose an algorithm for interactive ray‐casting of algebraic surfaces of high degree. A key point of our approach is a polynomial form adapted to the view frustum. This so called frustum form yields simple expressions for the Bernstein form of the ray equations, which can be computed efficiently using matrix products and pre‐computed quantities. Numerical root‐finding is performed using B‐spline and Bézier techniques, and we compare the performances of recent and classical algorithms. Furthermore, we propose a simple and fairly efficient anti‐aliasing scheme, based on a combination of screen space and object space techniques. We show how our algorithms can be implemented on streaming architectures with single precision, and demonstrate interactive frame‐rates for degrees up to 16.     

SATO: Surface Area Traversal Order for Shadow Ray Tracing

We present the surface area traversal order (SATO) metric to accelerate shadow ray traversal. Our formulation uses the surface area of each child node to compute the TO. In this metric, we give a traversal priority to the child node with the larger surface area to quickly find occluders. Our algorithm reduces the pre‐processing overhead significantly, and is much faster than other metrics. Overall, the SATO is useful for ray tracing large and complex dynamic scenes (e.g. a few million triangles) with shadows.     

Accelerating Ray Tracing using Constrained Tetrahedralizations

In this paper we introduce the constrained tetrahedralization as a new acceleration structure for ray tracing. A constrained tetrahedralization of a scene is a tetrahedralization that respects the faces of the scene geometry. The closest intersection of a ray with a scene is found by traversing this tetrahedralization along the ray, one tetrahedron at a time. We show that constrained tetrahedralizations are a viable alternative to current acceleration structures, and that they have a number of unique properties that set them apart from other acceleration structures: constrained tetrahedralizations are not hierarchical yet adaptive; the complexity of traversing them is a function of local geometric complexity rather than global geometric complexity; constrained tetrahedralizations support deforming geometry without any effort; and they have the potential to unify several data structures currently used in global illumination.     

On‐the‐fly generation and rendering of infinite cities on the GPU

In this paper, we present a new approach for shape‐grammar‐based generation and rendering of huge cities in real‐time on the graphics processing unit (GPU). Traditional approaches rely on evaluating a shape grammar and storing the geometry produced as a preprocessing step. During rendering, the pregenerated data is then streamed to the GPU. By interweaving generation and rendering, we overcome the problems and limitations of streaming pregenerated data. Using our methods of visibility pruning and adaptive level of detail, we are able to dynamically generate only the geometry needed to render the current view in real‐time directly on the GPU. We also present a robust and efficient way to dynamically update a scene's derivation tree and geometry, enabling us to exploit frame‐to‐frame coherence. Our combined generation and rendering is significantly faster than all previous work. For detailed scenes, we are capable of generating geometry more rapidly than even just copying pregenerated data from main memory, enabling us to render cities with thousands of buildings at up to 100 frames per second, even with the camera moving at supersonic speed.