SIGGRAPH 2010: Imaging of Interest
August 9, 2010 9:15 AM PT

Johannes Kopf and Michael Cohen recently had a busy week.

Kopf and Cohen, researchers within the Interactive Visual Media Group at Microsoft Research Redmond, joined teammate Richard Szeliski and Billy Chen of Virtual Earth Labs, in Los Angeles from July 25 to 29 for SIGGRAPH 2010, the 37th International Conference and Exhibition on Computer Graphics and Interactive Techniques.

Approximately 25,000 professionals in the fields of computer graphics and interactive technologies gathered at the Los Angeles Convention Center, and Kopf and colleagues attracted considerable interest for their SIGGRAPH-accepted paper Street Slide: Browsing Street Level Imagery.

That attention, Cohen suggests, constitutes the value that researchers gain from attending the event.

“SIGGRAPH provides two benefits,” Cohen says. “First, the competitive review process provides validation for the work and also encourages discipline in project development.

“Second, it provides a great showcase to launch the work and get feedback from colleagues.”

Microsoft Research participants in SIGGRAPH  this year have in the past contributed to the development of Photosynth, Image Composite Editor, Bing Maps Birds Eye View, the computer-graphics effects in the Halotrilogy, and Photo Fuse, now in Windows Live Photo Gallery.  Each is an illustration of how work in computer graphics and vision technology can be incorporated into products and services.

People and Pixels

The theme for SIGGRAPH 2010 was The People Behind the Pixels. The organizers stressed the personalities, the stories, and the history behind the key content presented during the conference.

Several of those people behind the pixels come from Microsoft Research, which presented 17 (13 percent) of the 133 papers accepted for presentation during the event.

Of those, Street Slide, a novel browsing interface for street-level imagery that combines the best aspects of the immersive nature of “bubbles” with the overview provided by multiperspective strip panoramas, received significant interest, in large part because of the ingenious way in which the researchers tackle problems with current street-side imagery on the web.

“Bubbles” are immersive, 360-degree panoramas—each from a particular perspective—that enable users to visit cities virtually, pan around to inspect the area within the bubble, and zoom in once something of interest is spotted.

But as breathtaking as such an experience can be, there are a couple of problems with the technology.

For one thing, it is not satisfying to jump from one bubble to the next. Users don’t get a good visual sense of chunks of territory larger than the bubbles—say, an entire city block, or an extended length of a long, busy street.

An alternative, multiperspective strip panorama, while able to deliver a visual summary of a city street, can lack the full realism of the immersive panorama.

“The primary goal,” Cohen says, “was to develop a new interface for street-level imagery that would allow simpler and more intuitive navigation along streets.

“The interface should also provide easy affordances for discovering local information.”

Multiperspective Street Slide panorama
A multiperspective Street Slide panorama with navigational aids and a mini-map.

Street Slide offers a variety of approaches for making the online navigation of city streets more seamless and realistic.

“The key contributions,” Cohen states, “include a means to align and combine imagery for multiple panoramas captured every few meters along a street. This construction is performed on the fly as a user pans along a street. The experience depicts parallax, providing a sense of immersion.

“Affordances for turning to the other side of the street or to turn onto cross streets create a simple navigation method. Information on what can be found along the streets is provided in the form of icons suspended below the street side.”

A user study undertaken as part of the project confirmed that the Street Slide interface greatly enhanced visual searching of street-side imagery, in large part because the technology incorporates and combines the best aspects of the immersive nature of bubbles with the overview provided by multiperspective strip panoramas.

Human Scale

“This work will provide a very familiar, human-scale experience of local areas,” Cohen concludes. “It overcomes the highly restrictive view of the street provided in current single-point panorama interfaces. The layout provides a more intuitive canvas for embedding information.”

Street Slide isn’t the only paper from the Interactive Visual Media Group accepted for SIGGRAPH 2010. Also included was GradientShop: A Gradient-Domain Optimization Framework for Image and Video Filtering, co-written by Cohen, along with Microsoft Research colleague Larry Zitnick, Pravin Bhat of Weta Digital, and Brian Curless of the University of Washington. And Szeliski also was represented by a second paper, Ambient Point Clouds for View Interpolation, co-written with Microsoft colleague Drew Steedly and five collaborators from Technische Universität Darmstadt: Michael Goesele, Jena Ackermann, Simon Fuhrmann, Carsten Haubold, and Ronny Klowsky.

The most active Microsoft researcher in the event, however, was Baining Guo, assistant managing director of Microsoft Research Asia and head of that facility’s Internet Graphics Group. Guo is listed as co-author of no fewer than six SIGGRAPH papers, on subjects ranging from rendering hair and scattering to fabrication, GPU rendering, textures, and appearance capture and image processing.

Microsoft Research personnel filled a number of other roles during the conference. Kurt Akeley, principal researcher at Microsoft Research Silicon Valley, helped present a course called Beyond Programmable Shading II, about using parallel programming techniques and architectures to create advanced interactive graphics algorithms. Bill Baxter, Nelson Chu, and Naga Govindaraju of the eXtreme Computing Group gave a talk on their Project Gustav: Immersive Digital Painting, a realistic painting-system prototype. And three representatives of Microsoft Research Redmond served as session chairs: John Snyder for the Architectural Patterns session, Szeliski for the Video session, and Charles Loop for Surface Fields.

Papers written all or in part by Microsoft Research and accepted for SIGGRAPH 2010:

Ambient Point Clouds for View Interpolation
Michael Goesele, Technische Universität Darmstadt; Jens Ackermann, Technische Universität Darmstadt; Simon Fuhrmann, Technische Universität Darmstadt; Carsten Haubold, Technische Universität Darmstadt; Ronny Klowsky, Technische Universität Darmstadt; Drew Steedly, Microsoft; and Richard Szeliski, Microsoft Research Redmond.

Distributed Gradient-Domain Processing of Planar and Spherical Images
Michael Kazhdan, Johns Hopkins University; Dinoj Surendran, Microsoft Research Redmond; and Hugues Hoppe, Microsoft Research Redmond.

Fabricating Spatially Varying Subsurface Scattering
Yue Dong, Tsinghua University; Jiaping Wang, Microsoft Research Asia; Fabio Pellacini, Dartmouth College; Xin Tong, Microsoft Research Asia; and Baining Guo, Microsoft Research Asia.

Face Poser: Interactive Modeling of 3D Facial Expressions Using Facial Priors
Manfred Lau, Carnegie Mellon University; Jinxiang Chai, Texas A&M University; Ying-Qing Xu, Microsoft Research Asia; and Heung-Yeung Shum, Microsoft Research Asia.

GradientShop: A Gradient-Domain Optimization Framework for Image and Video Filtering
Pravin Bhat, Weta Digital; C. Lawrence Zitnick, Microsoft Research Redmond; Michael F. Cohen, Microsoft Research Redmond; and Brian Curless, University of Washington.

Image Deblurring Using Inertial Measurement Sensors
Neel Joshi, Microsoft Research Redmond; Sing Bing Kang, Microsoft Research Redmond; C. Lawrence Zitnick, Microsoft Research Redmond; and Richard Szeliski, Microsoft Research Redmond.

Interactive Hair Rendering Under Environment Lighting
Zhong Ren, Microsoft Research Asia; Kun Zhou, Zhejiang University; Tengfei Li, Zhejiang University; Wei Hua, Zhejiang University; and Baining Guo, Microsoft Research Asia.

Line-Space Gathering for Single Scattering in Large Scenes
Xin Sun, Microsoft Research Asia; Kun Zhou, Zhejiang University; Stephen Lin, Microsoft Research Asia; and Baining Guo, Microsoft Research Asia.

Manifold Bootstrapping for SVBRDF Capture
Yue Dong, Tsinghua University; Jiaping Wang, Microsoft Research Asia; Xin Tong, Microsoft Research Asia; John Snyder, Microsoft Research Redmond; Yanxiang Lan, Tsinghua University; Moshe Ben-Ezra, Microsoft Research Asia; and Baining Guo, Microsoft Research Asia.

Micropolygon Ray Tracing with Defocus and Motion Blur
Qiming Hou, Tsinghua University; Hao Qin, Zhejiang University; Wenyao Li, Zhejiang University; Baining Guo, Microsoft Research Asia; and Kun Zhou, Zhejiang University.

Multi-Class Blue Noise Sampling
Li-Yi Wei, eXtreme Computing Group.

Personal Photo Enhancement Using Example Images
Neel Joshi, Microsoft Research Redmond; Wojciech Matusik, Disney Research; Edward H. Adelson, Massachusetts Institute of Technology; and David J. Kriegman, University of California, San Diego.

Precomputed Wave Simulation for Real-Time Sound Propagation of Dynamic Sources in Complex Scenes
Nikunj Raghuvanshi, eXtreme Computing Group; John Snyder, Microsoft Research Redmond; Ravish Mehra, University of North Carolina at Chapel Hill; Ming Lin, University of North Carolina at Chapel Hill; and Naga Govindaraju, eXtreme Computing Group.

Reducing Shading on GPUs Using Quad-Fragment Merging
Kayvon Fatahalian, Stanford University; Solomon Boulos, Stanford University; James Hegarty, Stanford University; Kurt Akeley, Microsoft Research Silicon Valley; William R. Mark, Intel; Henry Moreton, NVIDIA; and Pat Hanrahan, Stanford University.

Sampling-Based Contact-Rich Motion Control
Libin Liu, Tsinghua University; KangKang Yin, Microsoft Research Asia; Michiel van de Panne, The University of British Columbia; Tianjia Shao, Tsinghua University; and Weiwei Xu, Microsoft Research Asia.

Street Slide: Browsing Street-Level Imagery
Johannes Kopf, Microsoft Research Redmond; Billy Chen, Microsoft; Richard Szeliski, Microsoft Research Redmond; and Michael F. Cohen, Microsoft Research Redmond.

Vector Solid Textures
Lvdi Wang, Tsinghua University; Kun Zhou, Zhejiang University; Yizhou Yu, University of Illinois at Urbana-Champaign; and Baining Guo, Microsoft Research Asia.