Abstract:

We introduce image-space radiosity and a hierarchical variant as a method for interactively approximating diffuse indirect illumination in fully dynamic scenes. As oft observed, diffuse indirect illumination contains mainly low-frequency details that do not require independent computations at every pixel. Prior work leverages this to reduce computation costs by clustering and caching samples in world or object space. This often involves scene preprocessing, complex data structures for caching, or wasted computations outside the view frustum. We instead propose clustering computations in image space, allowing the use of cheap hardware mipmapping and implicit quadtrees to allow coarser illumination computations. We build on a recently introduced multiresolution splatting technique combined with an image-space lightcut algorithm to intelligently choose virtual point lights for an interactive, one-bounce instant radiosity solution. Intelligently selecting point lights from our reflective shadow map enables temporally coherent illumination similar to results using more than 4096 regularly-sampled VPLs.

Abstract:

Global illumination provides a visual richness not achievable with the direct illumination models used by most interactive applications. To generate global effects, numerous approximations attempt to reduce global illumination costs to levels feasible in interactive contexts. One such approximation, reflective shadow maps, samples a shadow map to identify secondary light sources whose contributions are splatted into eye-space. This splatting introduces significant overdraw that is usually reduced by artificially shrinking each splat's radius of influence. This paper introduces a new, multi-resolution approach for interactively splatting indirect illumination. Instead of reducing GPU fill rate by reducing splat size, we reduce fill rate by rendering splats into a multi-resolution buffer. This takes advantage of the low-frequency nature of diffuse and glossy indirect lighting, allowing rendering of indirect contributions at low resolution where lighting changes slowly and at high resolution near discontinuities. Because this multi-resolution rendering occurs on a per-splat basis, we can significantly reduce fill rate without arbitrarily clipping splat contributions below a given threshold---those regions simply are rendered at a coarse resolution.

Abstract:

Global illumination provides a visual richness not achievable with the direct illumination models used by most interactive applications. To generate global effects, numerous approximations attempt to reduce global illumination costs to levels feasible in interactive contexts. One such approximation, reflective shadow maps, samples a shadow map to identify secondary light sources whose contributions are splatted into eye-space. This splatting introduces significant overdraw that is usually reduced by artificially shrinking each splat's radius of influence. This paper introduces a new, multi-resolution approach for interactively splatting indirect illumination. Instead of reducing GPU fill rate by reducing splat size, we reduce fill rate by rendering splats into a multi-resolution buffer. This takes advantage of the low-frequency nature of diffuse and glossy indirect lighting, allowing rendering of indirect contributions at low resolution where lighting changes slowly and at high resolution near discontinuities. Because this multi-resolution rendering occurs on a per-splat basis, we can significantly reduce fill rate without arbitrarily clipping splat contributions below a given threshold---those regions simply are rendered at a coarse resolution.

Abstract:

Caustic maps provide an interactive image-space method to render caustics, the focusing of light via reflection and refraction. Unfortunately, caustic mapping suffers problems similar to shadow mapping: aliasing from poor sampling and map projection as well as temporal incoherency from frame-to-frame sampling variations. To reduce these problems, researchers have suggested methods ranging from caustic blurring to building a multiresolution caustic map. Yet these all require a fixed photon sampling, precluding the use of importance-based photon densities. This paper introduces adaptive caustic maps, which allow dynamic photon sampling using a hierarchical renderer enabled by a new deferred shading technique. Alone, deferred shading speeds rendering of refractive geometry up to 25% and with adaptive sampling speeds caustic rendering up to 200%. These benefits are particularly noticable for complex geometry or using millions of photons. While developed for a GPU rasterizer, adaptive caustic map creation can be performed by any renderer that individually traces photons, e.g., a GPU ray tracer.

Abstract:

Scattering effects arising from participating media, such as smoke, haze, and fog, dramatically add to perceived realism in renderings. As shadows affect illumination throughout an environment, they significantly diminish scattering effects in umbral regions. Unlike surface shadowing, accurate volumetric shadows require simultaneously integrating illumination, scattering, and attenuation throughout the volume, which proves challenging for interactive applications. We propose a method for rendering volumetric shadows in homogeneous single scattering media that combines shadow volume and ray marching techniques, which eliminates performance deficiencies inherent in both. We extend this approach to interactively render shadows from textured lights and show results under two scattering models.

Abstract:

Interactive applications typically rely on local models for lighting, occasionally augmented by GPU-friendly methods for approximating global illumination. Caustic mapping approximates the specular focusing of light using a light-space image, akin to a shadow map, which is projected onto the scene during final rendering. Unfortunately, existing caustic map implementations must choose between quality and speed. Quickly generated maps use few photons and look extremely blurry, while sharper maps created from millions of photons only render at a few frames per second. This paper introduces a number of hierarchical enhancements to caustic mapping that allow real-time rendering with high quality caustic maps, even when using maps from multiple light sources. These techniques utilize the geometry processing stage of recent GPUs to avoid processing every photon and to render a pyramidal caustic map that allows photon splats of varying diameters without the increased costs inherent in large splats.

Abstract:

Interactivity requires tradeoffs to achieve the right balance between rendering quality and speed. In practice, today's applications restrict lighting to mainly direct illumination, sometimes augmented by precomputed transfer techniques for diffuse global effects. Dynamic high-frequency specular effects, such as caustics, are largely lacking due to the high costs for recomputation each frame. Recent work has introduced a variety of related caustics approximations that interactively render light-space photons into a photon buffer, gather them into a caustic map, and project this map onto the scene similar to shadow mapping. While the process is simple and straightforward, the discretization of light into a finite number of uniformly-distributed photons leads to undersampling and aliasing artifacts. This paper examines two techniques for reducing these artifacts using varying sized photon splats. Conceptually, these are similar to the variable-radius $k$-nearest neighbor search used in photon mapping, allowing noise reduction in areas of low photon density while maintaining crisp caustics at focal points. Our techniques improve image quality at a modest cost that is significantly cheaper than supersampling the photon buffer.

Abstract:

A fundamental challenge for existing shadow map based algorithms is dealing with partially illuminated surfaces. A conventional shadow map built with a pinhole camera only provides a binary visibility sorting of the scene, and this all-or-nothing approach to visibility does not capture penumbral regions. We present an interactive soft shadow algorithm based on a variant of the depth discontinuity occlusion camera, a non-pinhole camera with rays that reach around blockers to sample normally hidden surfaces. Our soft shadow occlusion camera (SSOC) classifies a fragment on a continuum from fully visible to fully hidden, as seen from the light. The SSOC is used directly in fragment illumination computation without building an explicit ``soft shadow map.'' This method renders plausible soft shadows at interactive speeds under fully dynamic conditions.

Abstract:

A requirement for rendering realistic images interactively is efficiently simulating material properties. Recent techniques have improved the quality for interactively rendering dielectric materials, but have mostly neglected a phenomenon associated with refraction, namely, total internal reflection. We present an algorithm to approximate total internal reflection on commodity graphics hardware using a ray-depth map intersection technique that is interactive and requires no precomputation. Our results compare favorably with ray traced images and improve upon approaches that avoid total internal reflection.

Abstract:

Interactive applications require simplifications to lighting, geometry, and material properties that preclude many effects encountered in the physical world. Until recently only the most simplistic reflections and refractions could be performed interactively, but state-of-the-art research has lifted some restrictions on such materials. This paper builds upon this work, but examines reflection and refraction from the light's viewpoint to achieve interactive caustics from point sources. Our technique emits photons from the light and stores the results in image-space, similar to a shadow map. We then examine various techniques for gathering these photons, comparing their advantages and disadvantages for rendering caustics. These approaches run interactively on modern GPUs, work in conjunction with existing techniques for rendering specular materials, and produce images competitive with offline renderings using comparable numbers of photons.

(See similar, independent work here, here, and here.)

Abstract:

In many applications, volumetric datasets are examined by displaying isosurfaces, surfaces where data, or some function of the data, takes on a given value. Interactive applications typically use local lighting models to render such surfaces. This work introduces a method to precompute or lazily compute global illumination to improve interactive isosurface renderings. The precomputed illumination resides in a separate volume and includes direct light, shadows, and interreflections. Using this volume, interactive globally illuminated renderings of isosurfaces becomes feasible while still allowing dynamic manipulation of viewpoint and isovalue.

Abstract:

We present a mathematical framework for enforcing energy conservation in a BRDF by specifying halfway vector distributions in simple two-dimensional domains. Energy-conserving BRDFs can produce plausible rendered images with accurate reflectance behavior, especially near grazing angles. Using our framework, we create an empirical BRDF that allows easy specification of diffuse, specular, and retroreflective materials. We also present a second BRDF model that is useful for data fitting; although it does not preserve energy, it uses the same halfway vector domain as the first model. We show that this data-fitting BRDF can be used to match measured data extremely well using only a small set of parameters. We believe that this is an improvement over table-based lookups and factored versions of BRDF data.

Abstract:

Interactive applications often strive for realism, but framerate constraints usually limit realistic effects to those that run efficiently in graphics hardware. One effect largely ignored in interactive applications is refraction. We build upon a simple, image-space approach to refraction that easily runs on modern graphics cards. This image-space approach requires two passes on a GPU, and allows refraction of distant environments through two interfaces. Our works explores extensions allowing the refraction of nearby opaque objects, at the cost of one additional pass to render nearby geometry to texture and a more complex fragment shader for computing refracted color. Like all image-based algorithms, aliasing can occur in certain circumstances, especially when a few texels are magnified to cover a sizable portion of screen space. However, our plausible refractions should suffice for many applications.

Abstract:

Many interactive applications strive for realistic renderings, but framerate constraints usually limit realism to effects that run efficiently in graphics hardware. One effect largely ignored in such applications is refraction. We introduce a simple, image-space approach to refractions that easily runs on modern graphics cards. Our method requires two passes on a GPU, and allows refraction of a distant environment through two interfaces, compared to current interactive techniques that are restricted to a single interface. Like all image-based algorithms, aliasing can occur in certain circumstances, but the plausible refractions generated with our approach should suffice for many applications.

Abstract:

Bright patterns of light focused via reflective or refractive objects onto matte surfaces are called ``caustics''. We present a method for rendering dynamic scenes with moving caustics at interactive rates. This technique requires some simplifying assumptions about caustic behavior allowing us to consider it a local spatial property which we sample in a pre-processing stage. Storing the caustic locally limits caustic rendering to a simple lookup. We examine a number of ways to represent this data, allowing us to trade between accuracy, storage, run time, and precomputation time.

Abstract:

Interactive global illumination remains an elusive goal in rendering, as energy from every portion of the scene contributes to the final image. Integrating over a complex scene, with a polygon count in the millions or more, proves difficulty even for static techniques. Interactive with such complex environments while maintaining high quality rendering generally requires recomputing the paths of countless photons using a small number of CPUs.

This dissertation examines a simplified approach to interactive global illumination. Observing that local illumination computations can be performed interactively even on fairly simple graphics accelerators, a reduction of global illumination problems to local problems would allow interactive rendering. A number of techniques are suggested that simplify global illumination to specific global illumination effects (e.g., diffuse interreflection, soft shadows, and caustics), which can individually be sampled at a local level. Rendering these simplified global illumination effects reduces to a few lookups, which can easily be done at interactive rates. While some tradeoffs exist between rendering speed, rendering quality, and memory consumption, these techniques show that approximating global illumination locally allows interactivity while still maintaining significant realism.

Abstract:

Generating soft shadows quickly is difficult. Few techniques have enough flexibility to interactively render soft shadows in scenes with arbitrarily complex occluders and receivers. This paper introduces the penumbra map, which extends current shadow map techniques to interactively approximate soft shadows. Using object silhouette edges, as seen from the center of an area light, a map is generated containing approximate penumbral regions. Rendering requires two lookups, one into each the penumbra and shadow maps. Penumbra maps allow arbitrary dynamic models to easily shadow themselves and other nearby complex objects with plausible penumbrae.

Abstract:

In computer graphics, bright patterns of light focused onto matte surfaces are called ``caustics''. We present a method for rendering dynamic scenes with moving caustics at interactive rates. This technique requires some simplifying assumptions about caustic behavior allowing us to consider it a local spatial property which we sample in a pre-processing stage. Storing the caustic locally limits caustic rendering to a simple lookup. We examine a number of ways to represent this data, allowing us to trade between accuracy, storage, run time, and precomputation time.