MANHASSET, N.Y. The leading name in computing graphics these days isn't Pixar or Nvidia or 3Dlabs. It's Microsoft Corp. With the help of sophisticated algorithms and computing techniques, the company is finding new ways to manipulate pixels and re-create reality.
A testament to its efforts was the awarding of the prestigious Computer Graphics Achievement Award to senior researcher Hugues Hoppe of Microsoft Research last week at Siggraph 2004, the computer graphics industry's biggest event.
Hoppe was the eighth Microsoft Research employee to win the award since 1983.
Hoppe received the achievement award for his pioneering work on surface reconstruction, progressive meshes, geometry texturing and geometry images. His seminal paper on progressive meshes is one of the most widely cited and influential papers in computer graphics. Mesh parameterization is essential in mapping texture images onto surfaces, greatly enhancing their visual detail and quality.
At the 31st incarnation of Siggraph, held in Los Angeles, Microsoft presented 12 research papers, nine of them done in partnership with universities around the world. Microsoft Research's contributions accounted for nearly 15 percent of the conference's papers, more than any other single organization.
Computer scientists from Microsoft Research's labs in Beijing; Cambridge, England; and Redmond, Wash., collaborated on the papers with colleagues and affiliates from the University of Washington, University of Utah, Stanford University, Hong Kong University of Science and Technology, University of Toronto, University of Illinois at Urbana-Champaign, Zhejiang University and the Institute of Computing Technology, Chinese Academy of Sciences.
Technologies and projects range from investigating novel methods of video-based rendering to pursuing innovative advancements in interactive modeling, 3-D textures, digital photography and large meshes and graphics-processing unit programming.
Microsoft Research has attracted scientists with a passion for computer graphics. Founded in 1991, it does both basic and applied research, not only in computer graphics, but also in computer science and software engineering. Among its goals: simplifying and enhancing the computing experience, reducing the cost of writing and maintaining software, and facilitating the creation of new types of software. Microsoft Research employs more than 700 people at five labs on three continents, focused on more than 55 areas of computing.
Hoppe has been a visible and prolific researcher for 10 years, having worked at Microsoft Research for his entire professional career. Several of his papers are considered milestones that have significantly influenced the computer graphics field.
With eight Microsoft researchers now holding the Computer Graphics Achievement Award, another recipient, industry veteran Kurt Akeley, will join the lab this fall. Akeley, a co-founder of Silicon Graphics Inc., has been a leading contributor to the architecture, design and realization of high-performance 3-D graphics hardware systems, including GTX, VGX, OpenGL and RealityEngine.
Among the papers that Microsoft presented at Siggraph 2004 was a detailed look at a new algorithm that automatically extracts accurate geometry information so that a user can create special visual effects, such as the now-famous Matrix freeze-frame effect, right on a desktop on demand and cost-effectively.
Another paper described a technique that greatly reduces the amount of run-time processing by modeling an object from a small surface sample whose light-scattering properties have been previously simulated and stored. Thus, complex objects can be vividly rendered hundreds of times faster than before by synthesizing the properties of this surface sample over an object.
Award recipient Hoppe either wrote or collaborated with others on three of the papers that were presented at this year's Siggraph. According to an account on the Microsoft Research Web site, Hoppe had a ground-breaking beginning at the lab. Just starting his career, he had an interesting problem to solve. A group from Stanford University, which had visited Florence, Italy, to scan Michelangelo's David, needed meshes with billions of triangles to reconstruct the David digitally. Hoppe's effort to simplify their task led to his pioneering computer graphics work in progressive meshes.
He simplified the geometric model by reducing the number of triangles needed, thereby improving the speed of rendering. By simplifying the graphic models once, then storing the history of the simplification, all the tiny perturbations to the mesh at each rendering step could be recorded and played back to reintroduce as much detail as needed.
His subsequent work on progressive meshes, called geomorphing, allows for smooth transitions as the level of detail changes, eliminating a "jumping" image as the object is rendered.
Of the three papers Hoppe presented at Siggraph he considers his Geometry ClipMaps most significant.
According to the description of the paper, Hoppe has produced a giant grid of the United States using elevation data sampled every 30 meters. His challenge: how to convey this geography to the user almost instantaneously. He and his collaborators came up with a way to use regular grids that vary with the distance from the viewer. The grids mirror real-world geometry. That is, things that are farther away have the same size in screen space as they would in real-world space. He explained that "as the viewer 'flies' around the world, the grids move and adapt the meshes to the distance from the viewer."
He attributes the technique to today's faster hardware graphics. Hoppe's method eliminated triangulation at every frame, and had the more powerful hardware use a "brute-force" approach on the entire world, eliminating the need to fine-tune every local terrain. Hoppe's grid work is different from his progressive-mesh contributions and reflects the impact that computer graphics hardware has had on the industry. While the main processor once did more of the work on highly irregular meshes, today a powerful graphics processor creates regular grids on the fly.
In another Siggraph paper on "Digital Photography with Flash and Non-Flash Pairs," Hoppe describes how to achieve a sharper image by combining the best parts of a picture taken with a flash and one taken without. The technique also reduces the red-eye effect common to pictures taken with a flash.
When a flash picture is taken, most often only the eye turns red, so Hoppe's program identifies areas that appear dark in the nonflash picture and red in the flash picture. By integrating the nonflash eye into the flash version he automatically corrects red-eye.
Over the years, Hoppe has co-authored 20 papers presented at Siggraph conferences. He is also a professor at the University of Washington, where he collaborates with another recipient of this year's Siggraph's awards, Zoran Popovi'c.
A researcher at the University of Washington, Popovi'c earned the Significant New Researcher Award.
Popovi'c co-authored a paper this year on fluid simulations, a gnawing challenge to computer graphics specialists. His paper describes a way to control fluid simulations through gradient-based nonlinear optimization using an adjoint technique. Detailed control of hundreds of thousands of free variables for simulating a fluid has been achieved up to now only at relatively coarse scales. In their work, Popovi'c and his co-authors showed how the adjoint method drastically reduces the simulation system's dependence on the number of control parameters, making the high-level control of free-surface liquids possible for the first time.