Physics-Based Glossy Reflection: This image is from Jim Arvo's paper
Applications of Irradiance Tensors....
The image depicts a "glossy" or imperfect reflection of a collection of polygons.
The reflection was computed analytically using a technique described
in the paper.
Empirical Glossy Reflection: These images were rendered using the multi-pass pipeline techniques described in Multi-Pass Pipeline Rendering....
They simulate the physics-based rendering depicted above, and required under .5 seconds of rendering time.
Physics-Based Frosted Glass: These images are from Jim Arvo's paper Applications of Irradiance Tensors.... They depict a polygonal butterfly behind two different pieces of frosted glass. The transmitted light was computed analytically at each pixel.
Empirical Frosted Glass: These images are computed using multi-pass pipeline rendering described in Multi-Pass Pipeline Rendering.... They simulate the physics-based renderings above, but required less than 0.1 seconds to render.
Light Filtering: These images demonstrate filtering of light and shadows through bevelled glass.
Recursive Image Iterations: This series of images demonstrate the multiple passes made using our techniques.
Specular Bathroom: These images were generated usinng multi-pass pipeline rendering techniques described in Multi-Pass Pipeline Rendering.... They include specular lighting effects, reflection, refraction, and filtered lighting.
Radiance vs. Pipeline Image: These images demonstrate a comparison of an image generated with Greg Ward's Radiance system (left) versus a multi-pass pipeline generated image (right).
Adding Indirect Illumination: These images demonstrate using the indirect component (left) generated from a radiosity solution as the first pass in our multi-pass pipeline rendering techniques, which combined with the direct effects (middle), produce full scene illumination (right).
Varying Shadow Settings: Image 1 (3.1 sec) shows the lighting passes for a single lightsource with one reflective and one refractive suface. In Image 2 (2.7 sec), shadow volumes which encounter more than one specular surface are eliminated. This effect is apparent in the shadow pattern on the tub. Image 3 (2.1 sec) shows the original situation without reflected and refracted shadows (obects before the specular surface). Note the light volume is larger because the frame no longer blocks light from reaching the mirror, and the shower door handle no longer casts a shadow through the door. Image 4 (1.8) replaces the refractive shower door with a purely transparent, non-refracting surface. Finally, Image 5 (2.6 sec) demonstrates the original scene with a shadow recursion depth of one instead of two.
Varying Specular Settings: Image 1 (63.5 sec) shows three specular surfaces (mirror, shower door, and floor) at a specular depth of two, and Image 2 (28.0 sec) at a specular depth of one. In Image 3 (11.4 sec), the partial specular floor image has been excluded at a depth of two. This is then repeated at a depth of one in Image 4 (6.9 sec). In Image 5 (3.4 sec), the refractive shower door is replaced with a non-refractive transparent surface at a depth of two.
Dynamic Scene Rendering: These images are from an animation demonstrating our techniques in a dynamic environment. The scene has 3 specular surfaces (2 mirrors, 1 glass door), with 4 lights. Only shadows and lighting effects which are effected by the moving objects are recomputed for each frame. Frame rendering times range from 15-65 seconds.
