| Animations | Bio | Home | Images | Links | Publications |
Global illumination is a superset of radiosity and ray tracing. The goal is to compute all possible light interactions in a given scene, and thus obtain a truely photorealistic image. All combinations of diffuse and specular reflections and transmissions must be accounted for. Effects such as colour bleeding and caustics must be included in a global illumination simulation.
Most of the images here have been rendered using photon mapping. Others have been rendered using techniques described in the papers on my papers web-page.
| The Courtyard House with Curved Elements |
 |
This image from the animation "The Light of Mies van der Rohe" demonstrates how photon mapping can be used compute global illumination in a complex model. The sun is the only light source and most of the light in the image is due to indirect illumination. The animation can be found on the animations web-page. |
| Teapot illuminated by a high dynamic range environment |
 |
 |
 |
 |
|
|
|
|
|
The classic teapot illuminated by a high dynamic range environment (grace cathedral lightprobe by Paul Debevec). Environments such as this provide an easy way to achieve realistic lighting. With complex environment maps it pays to use structured importance sampling - except for specular materials.
| Cornell box replica - rendered using photon mapping |
 |
This is a replica of (one of) the cornell box model first used at the graphics lab. at Cornell. It is a typical radiosity scene with polygonal faces and diffuse materials. Here it is rendered using photon mapping. There is another page with more examples of Cornell box images. |
| A daylight simulation of Little Matterhorn |
 |
 | |
|
|
|
This is a daylight simulation of Little Matterhorn. Just 100,000 photons were used to simulate indirect illumination of the landscape. For the cloudy day, I used a simple procedural technique to create a layer of true volumetric clouds that would cover the landscape and still provide sufficient detail. Note the shadows on the landscape from the clouds and how the tip of Little Matterhorn sticks into a cloud. The model of Little Matterhorn was created at University of Utah. There is an animation of the clouds forming over the landscape on the animations web-page.
| A Rendering of the Full Moon |
 |
This image of the full moon as seen through a thin layer of clouds was rendered using the techniques from a paper describing a physically-based night sky model. |
| Mostly indirect illumination |
 |
This simple box model has mostly indirect illumination and many global illumination algorithms have trouble rendering the indirect illumination (through the glass windows) accurately. Photon mapping works quite well in this case where the initial sampling of photons from the lights provides a very powerful estimate for the rendering phase. The model was made by Christophe Valtin. |
| A diffuse Jaguar model rendered with pathtracing |
 |
This is a simple diffuse model of a Jaguar with 140000 triangles, and a big diffuse skydome. It was rendered with pathtracing in 1024x512 in 20 min. on a PIII-500. This is one of the few cases were path tracing is resonable fast. However, eliminating the last noise would probably require several orders of magnitude more computation. Today, this type of rendering with a big white skydome light source has been popularized under the name of ambient occlusion. |
| Ray Tracing, Photon Mapping, and Global Illumination |
|
|
|
|
|
|
1.5 seconds |
7 seconds |
12 seconds |
15 seconds |
|
Cornell box series showing the effect of simulating global illumination. All the images have been rendered at 1024x768 with 4 samples per pixel on a Dual P3 800MHz Linux-PC. The ray tracing version took 1.5 seconds, The version with soft shadows took 7 seconds. Adding caustics increased the rendering time to 10 seconds and 2 seconds of photon tracing. The global illumination version took 12 seconds to render and 3 seconds of photon tracing time. For this scene the global photon map has 200,000 photons and the caustics photon map has 50,000 photons.
Finally, the path tracing reference image on the right
took a long time to render. The important thing to notice is that it
looks the same as the version rendered using photon mapping.
| A Museum |
 |
The museum image was rendered using the photon map based two-pass global illumination method. It took 45 minutes to render on a 100Mhz Pentium. This was one of my early images using the full two-pass algorithm. Note the anisotropic shading of the cylinders as well as the caustic on the non-diffuse teapot from the anisotropic cylinder. |
| A Metalring |
 |
Try illuminating a metalring on a table and you will see a nice cardioid caustic on the table. This shape is easy to describe mathematically and it is often used to illustrate caustics created as light from the light sources is reflected or transmitted via specular surfaces. See the caustics images for more examples on caustics. |
| A Desk |
 |
This is a path tracing rendering of a simple desk scene. The image was rendered using 2000 samples per pixel, and in order to render it in finite time I ran a parallelized version of my renderer on 30 SGI workstations (MIPS R3000) - still the rendering time was roughly 30 hours! Notice that no importance sampling was used here as described in the Rendering Workshop 95 paper - this is the reference image. |
| Soft shadows and color bleeding simulated using Monte Carlo ray tracing (1993) |
 |
This image demonstrates the quality of soft shadows and interreflection that is obtained using Monte Carlo ray tracing. |
| Soft shadows and color bleeding simulated using finite elements (the radiosity algorithm) (1992) |
 |
This image illustrates the quality of soft shadows and interreflection that is obtained by using the Radiosity algorithm with adaptive substructuring. |