Microsoft Researchers Present Next-Gen Graphics and Software Tools at SIGGRAPH
If you want to glimpse the future of gaming and entertainment software, check out the presentations at the world’s largest and most prestigious graphics conference.
This week, at the annual SIGGRAPH meeting (in New Orleans this year), researchers from Microsoft and elsewhere are presenting cutting-edge techniques in computer animation, image and video processing, interactive interfaces, and digital special effects. Microsoft Research has had a strong presence at this conference for the past decade.
What’s particularly cool about SIGGRAPH is its combination of academic rigor and almost immediate commercial applications for games and movies. It’s the nature of the ultra-competitive entertainment industry: everyone wants to use this stuff as soon as it’s ready for prime time (and sometimes even when it isn’t).
Here are a few of the Microsoft projects that stood out to me:
—Sometimes it’s all about the hair. But in the digital world of games and special effects, hair on characters can be a big pain. Artists and programmers have to manually create hairstyles for each character or base them on photos of real hair, and these are costly and time-consuming processes. Now Baining Guo of Microsoft Research Asia has teamed up with researchers at the University of Illinois, Urbana-Champaign, as well as Tsinghua University and Zhejiang University in China, to create a new method for synthesizing digital hair (see image above). The software designs each hairdo at two levels—keeping track of 2-D and 3-D geometry—and uses a “hair clustering algorithm” to take an input hairdo and create a new, original ‘do’ that can be put on a new character or avatar. This has clear applications for crowd simulations, game development, and online virtual worlds.
—Now you can search for clouds in the cloud. If you do an image search on Google or Bing, the results are tied to text surrounding the images, not the content of the pictures itself. You could potentially get much better results if the software could find specific features in the image you’re looking for. Now Jian Sun of Microsoft Research Asia and his collaborators at Hong Kong University of Science and Technology and Beihang University have solved this problem for a certain class of images (see left): a collection of more than half a million photos of the sky. Their software, called SkyFinder, can take a search query like “landscape at sunset with the sun on the bottom left,” and returns all pictures fitting that description. The tool could be useful in photo editing and film production, to swap in different sky backgrounds for a particular shot or scene.
—There’s a new model for computer architecture. Microsoft graphics guru Kurt Akeley, together with Stanford University researchers including Pat Hanrahan (known for his work with Pixar, and co-founder of Seattle’s Tableau Software), have developed a new programming model and toolkit for future graphics processing units (GPUs). GPUs are specialized chips, found in most personal computers, mobile phones, and game consoles, that are designed to manipulate and render 3D graphics efficiently. In the past few years, though, GPUs have become powerful enough to do more general kinds of computations, and influence the way computing is done in general. (One example is Portland, OR-based Elemental Technologies, which takes advantage of GPUs to do advanced video encoding and processing for businesses.) Akeley and his collaborators are aiming for the day when GPUs redefine computer architecture; their toolkit could have applications not only in entertainment (for rendering sophisticated new graphics) but for the broader future of software development.