Research


My main interest is in computer graphics, especially ray tracing and global illumination. But I have also done research in computer vision and VLSI design.

I'm working at Pixar's office in Seattle. We're developing new features for Pixar's RenderMan renderer. RenderMan is used for rendering CG movies such as Toy Story, Cars, and Tintin and for special effects in movies such as Terminator 2, Jurassic Park, the newer Star Wars episodes, Harry Potter, Lord of the Rings, Avatar, Jungle Book, and many more.

One of the interesting research topics for PRMan is efficient computation of ray tracing and global illumination in extremely complex scenes. We've introduced a multiresolution geometry cache for efficient ray tracing in scenes with very complex geometry. We've also introduced the brick map format, an efficient, tiled 3D MIP map representation of surface textures and volume data. The brick map format is designed to work well with a multiresolution cache. Among their many uses, we use brick maps to represent global illumination in a radiosity atlas (or irradiance atlas), and brick maps can be used as geometric primitives with built-in level-of-detail. I'm also doing some work on point-based computations such as approximate occlusion, approximate color bleeding, and subsurface scattering. The goal is to make global illumination a useful tool for daily use in movie production.

Before Pixar, I worked in the R&D department at Square USA in Honolulu. Square USA created the Final Fantasy movie. I was mainly working on developing and optimizing the global illumination parts of the massively parallel in-house renderer Kilauea. The work focused on efficient simulation of global illumination in extremely complex scenes using a fully parallel and distributed version of the photon map method.

Prior to that, I worked at Mental Images in Berlin. Mental Image's main product is the ray tracer Mental Ray (used by IBM/Dassault Systemes, Softimage, Alias, and others). My research included efficient global illumination of complex scenes and participating media using the photon map method developed by Henrik Wann Jensen (now at UC San Diego). We proved that the photon map method can be used in a production-strength commercial renderer. We also extended the photon map method to handle participating media by introducing a volume photon map. The resulting method is fast and simple, but nevertheless general enough to handle nonhomogeneous media and anisotropic scattering. The method can efficiently simulate effects such as multiple volume scattering, color bleeding between volumes and surfaces, and volume caustics. I also developed a shader interface for non-photorealistic contour rendering.

I am fascinated by the use of importance to make rendering more efficient. Here's a separate web page about importance in rendering.

At the University of Washington I also did research in global illumination, but using finite element methods. Most of my work was extending methods known from diffuse global illumination ("radiosity") to general global illumination. We showed how the ideas of importance-based transport and refinement, wavelet analysis, and clustering can be combined to provide an efficient solution to the radiance transport problem. Importance is used to focus the computation on the interactions having the greatest impact on the final visible solution. Wavelets are used to provide an efficient method for representing radiance, importance, and the transport operator itself. Clustering enables us to simplify light and importance transport between distant objects. My two advisors were David Salesin (now at Adobe Systems) and Tony Derose (now at Pixar). In addition, I was fortunate to work with Eric Stollnitz (then a grad student at the Department of Applied Mathematics, now at Microsoft) and Dani Lischinski (who had a postdoc position at the time, now at the Hebrew University in Jerusalem). If you're interested in current graphics research at University of Washington, click here.

I also worked with computer vision at the University of Washington -- mostly color photometric stereo. It is an extension of Woodham's photometric stereo method, and utilizes the different colors of specular and diffuse reflection to further constrain the direction of the normals. This gives higher accuracy in the determined shape. My advisor for that work was Linda Shapiro.

In a previous life, I did some research with Henrik Hulgaard on automated synthesis of asynchronous, delay-insensitive VLSI chips. This research was done at the Technical University of Denmark with Jørgen Staunstrup as advisor. Delay-insensitive circuits work without a clock and regardless of delays on wires. This has the advantage that a result can be used as soon as it is computed (rather than having to wait until the next clock cycle), and that the large areas that are occupied by clock distribution on traditional chips is eliminated. The downside is that more signal wires are necessary. This seemed like a promising idea 15 years ago, and it still does today. Our contribution was a translator from a software description of an algorithm to a design implementing the algorithm as a delay-insensitive VLSI chip.


I regularly review articles for the SIGGRAPH, Eurographics, Graphics Interface, and Interactive 3D conferences, for the Eurographics Rendering Workshops/Symposia, and for ACM Transactions on Graphics, Computer Graphics Forum, Computers & Graphics, IEEE Transactions on Visualization and Computer Graphics, and the Journal of Graphics Tools. I've been on the Eurographics Rendering Workshop/Symposium program committee nearly every year since 1999 and was co-chair with Daniel Cohen-Or in 2003. I am on the program committee for the IEEE Symposia on Interactive Ray Tracing (now merged with High Performance Graphics) and was co-chair with Alexander Keller for the 2007 symposium. I've been on the SIGGRAPH papers committee as well.

I've had the honor of serving on the Ph.D. committee of two outstanding researchers: Frank Suykens (KU Leuven, 2002) and Wojciech Jarosz (UC San Diego, 2008). I've also been judging the graphics course rendering competitions at Stanford (2007), UC San Diego (2007), and ETH Zurich (2013).


My Erdös number is 3: Paul Erdös (0) — John Conway (1) — Tom Duff (2) — me (3).

Back to Per's home page.