February 2010

You are currently browsing the monthly archive for February 2010.

Ken Torrance, 69, passed away unexpectedly February 15. His last name should be familiar to many of you: Torrance-Sparrow and Cook-Torrance?are two reflection models that bear it.

He was a professor in the mechanical and aerospace engineering department at Cornell. Don Greenberg, head of the Program of Computer Graphics at Cornell, was his long-time collaborator and was one person who spoke today at a gathering in remembrance of Ken. He talked about how Dr. Torrance was not a man who blew his own horn, how he was humble about his contributions. The main thing for Ken was that he loved to teach. Some early accomplishments Don noted (with my own additions):

  • Published in 1967, Torrance-Sparrow was based on his PhD work (Sparrow was his advisor). It is a reflection model that was based on experimental data combined with theory. Microfacet theory had been around since the 1920’s, but had problems, such as going to infinity at an incidence angle of 90 degrees. Torrance-Sparrow describes off-specular peaks, i.e., the phenomenon where the brightest reflection off of roughened metal is not in the reflection direction. It’s worth noting that Blinn’s reflection model is directly derived from this work, putting it into vector form.
  • Cook-Torrance, in 1981, should be more familiar to computer graphics people, coalescing ideas about microfacet distribution and Fresnel reflectance and bringing in the idea of wavelength dependence. Rob Cook is currently the VP of Software Engineering at Pixar.
  • The idea of using heat transfer theory in computer graphics is Torrance’s idea, along with Greenberg; another great example of how cross-pollination between fields can yield fantastic results. The paper by Goral et al. (including Ken) in 1984 describes computing diffuse-diffuse interaction. It introduced two concepts to the field that have become key elements of global illumination theory and practice: form factors and radiosity. What’s interesting to me in this paper is that it includes something somewhat rare in our field, an explanation of an experimental setup to compare computer-generated results with reality. It’s the paper where the Cornell Cube was first used (empty! It made visibility computations ignorable). If you want to try a form of this experiment, look here.
  • I could go on and on, the themes being simulating light transport and materials along with physically measuring BRDFs and illumination spectral distributions—theory and experiment. You can see some other papers listed here. Here’s?the measurements lab in 1995.?SIGGRAPH recognized his accomplishments in 1994, giving him the Computer Graphics Achievement Award that year.

I had a passing acquaintance with Professor Torrance over the years. I was working on ray-traced shadows, not radiosity, during my time at the Cornell PCG lab from 1983-85. I recall seeing a photo on a monitor in the lab one night and asking Cindy Goral about it, why it was scanned in. She told me it was an image she had generated with radiosity techniques, which was utterly flabbergasting to me. Dr. Torrance was a smart guy and a nice person, but I didn’t know him much beyond that. All I knew was that he was helping Cindy, and in his life many other students, make astounding images.

If you have a personal memory of Ken, you may wish to leave a note in the family’s guestbook.

Tags: ,

This I3D 2010 paper by Anton Kaplanyan (Crytek) and Carsten Daschbacher (University of Stuttgart) is now online on Crytek’s publications page, with video and talk slides.? The paper extends and describes in more detail the real-time global illumination technique Anton presented at SIGGRAPH 2009.? The most significant extension over the SIGGRAPH 2009 presentation is the addition of coarse occlusion for indirect bounce light.

Although there have been many recent papers on “real-time” global illumination, this technique is the only one so far that is at all feasible for current-generation consoles.? Crytek’s new CryEngine 3 has implementations of it on XBox 360 and Playstation 3, and the technique will presumably be used in the upcoming Crysis 2 game as well.

Tags: , ,

Game Engine Gems

A little under a year ago, we mentioned the call for papers for a new “gems” book: Game Engine Gems, edited by Eric Lengyel.? It’s available for pre-order on Amazon, not surprising since it’s due to be launched at GDC, just two weeks from now.? The table of contents is available on the book website, and looks promising.? Although the book contains some rendering chapters, its focus is broader, similar to the successful Game Programming Gems series to which it is likely to be a worthy competitor.? I’ve added it to our upcoming books list.


I asked Daniel Scherzer about my post about his book. He said it’s about right (and the long subtitle is indeed a Verlag decision).

The cool thing that turned up: their upcoming STAR survey on hard shadows will be more theoretical and detailed than his thesis’ survey. It will be similar to the hard shadow section in the SIGGRAPH Asia 2009 course?Casting Shadows in Real Time (which has a solid 90 pages on direct illumination shadow algorithms, plus more on related methods).

I’ve updated the original post with this information, but wanted to make a separate post so that people wouldn’t miss this valuable overview on shadows and ambient occlusion (just a tiny bit on the latter).

Tags: , ,

GPU Pro 2 book CFP

GPU Pro (the rebranded ShaderX series) will be published soon. Wolfgang Engel, the editor in chief of the series, has now put out a call for participation for GPU Pro 2;?read about it here. Proposal deadline is May 17th.

Tags: ,

Amazon sent an auto-recommended of this book to me. Unlike last time, which was humorous but unrelated, I actually appreciate this one: “Temporal Coherence in Real-Time Rendering: Practical Approaches for Capitalizing on Temporal Coherence in the Domain of Real-Time Rendering,” by Daniel Scherzer.

At $81 for a 132 page book, I suspected it was a thesis reprint. Indeed it is: you can download the thesis from here. The thesis is 130 pages long, so my guess is the book form adds nothing (and subtracts $81).

So, you can download it for free now, but should you read it? Well, it is a thesis, which means it collects various papers and presents each in turn. This thesis focuses on using temporal coherence, i.e. use previous frames’ computations in various ways. It includes Daniel’s hard shadow (history buffer), soft shadow, and discrete LOD blending work, as listed here. Since it’s a thesis, the author can stretch out a bit more and cover various areas in depth. The focus is on improving image quality: hard shadows are higher resolution, soft shadows look smoother. There are limitations to his approaches, e.g., the lights are fixed in place, and objects generally should be static.

As with most theses, it also includes an extensive “previous work” section at the beginning. There is a 23 page overview of a number of shadow techniques and LOD work, explaining strengths and weaknesses. From my skim, this looks quite good; not quite all-encompassing (which is good: there are way too many shadow papers), but hitting most of the major areas of research. Let’s put it this way: if and when we write a fourth edition, I’ll certainly carefully read his categorization of various problems and think about how to integrate it into our section on shadows. His is the best recent overview of the subject that I’ve seen. He’s also the coauthor of an upcoming survey on hard shadows, not yet available for download but which I suspect is similar to his thesis’ overview.

Update: I asked Daniel about this post, he said it’s about right (and the long subtitle is indeed a Verlag decision). The book version contains an index, and different (non-copyright-protected) images. Also of interest, their upcoming STAR survey on hard shadows will be more theoretical and detailed, similar to the hard shadow section in the SIGGRAPH Asia 2009 course Casting Shadows in Real Time (which has a solid 90 pages on shadow algorithms).

Tags: , ,

Eric Preisz has a book coming out in time for GDC, “Video Game Optimization.” I haven’t seen it yet, but judging by his article on optimization on Gamasutra, it should be pretty good—he knows what he’s talking about.

By the way, assuming you’re using Google Chrome for browsing (it’s what the cool kids use), I found AutoPager Chrome to be a nice little extension. Instead of needing to click at the end of every page of an article, it glues page after page into one long scroll.

Tags: ,


SBIM = Sketch-Based Interfaces and Modeling workshop
NPAR = Symposium on Non-Photorealistic Animation and Rendering
CFP = Call for Participation

SBIM-NPAR 2010 will be held June 7-10 in Annecy, France. The NPAR call for participation is up; due date is March 15th, but April 2nd for any SIGGRAPH rejections to be considered. In addition to research, they’re specifically looking to the games community for production papers on how game developers create their GUIs, editing tools, and NPR effects.

Every other year NPAR is held in Annecy, and it sounds lovely. In the evenings the Annecy International Animated Film Festival takes place—how perfect! Someday I’ll figure out a way to attend; until then, I can but dream.

Tags: , , ,

I mentioned this I3D 2010 paper in a blog post a while ago, but there was no preprint for it at the time.? Now the floodgates of information have been opened with the preprint, a video, and an extended paper all available on the first author’s website.? The source code has also been open-sourced with the Apache 2.0 license and posted on Google Code.

I’m not that familiar with previous work in the area, so I don’t know how it stacks up against e.g. Gigavoxels; if anyone has any insight please let me know.? One interesting detail from the paper was the way they enhance each cubical voxel with a pair of bounding planes, which they call contours.

They show that the data structure enables casting rays significantly faster than against a triangle model of equivalent complexity.? Unfortunately, it appears to still be slower than rendering the triangles the “old-fashioned way”.? The technique also requires post-filtering since it does not allow for filtering of color and normal information (which is effectively point-sampled).? Finally, building the data structure takes a fair bit of time, making this not particularly well-suited for dynamic scenes.

Tags: ,

Made me laugh

I noticed Kirk & Hwu’s “Programming Massively Parallel Processors” book is back in stock at Amazon (and now with 2 mixed reviews), after being unavailable for a number of days. The part that made me laugh is Amazon’s ranking listing:

#1 in  Books >?Computers & Internet >?Hardware >?Mainframes & Minicomputers

If GPU computing isn’t the antithesis of mainframes and minicomputers, I’m not sure what is…

Tags: ,

« Older entries