Recent Advances in Digital Processing of Images and Audio

09/29/2006 4:00 32-123Henrique Malvar, PhD '86, Director, Microsoft ResearchDescription: Henrique Malvar hopes to impress a tough crowd with -tricks" hot from Microsoft's research labs. His show-and-tell features the latest magic in sound and image digital signal processing, soon to appear on your PC or laptop.

Malvar's researchers have improved computer microphones so they can isolate a human voice within an acoustic din. They've also managed to reduce the echo often encountered on cell phone calls, and to clean up the -mooshy" sound of speaker phones. Another nifty advance from Malvar's folks: eliminating transmission delays in voice over IP, by speeding up or slowing down incoming data packets.

Malvar's got high expectations for breakthroughs from the image processing labs. -I hope these will affect everybody's lives," he says. Malvar unveiled -image stitching" software that can automatically create panoramas from a bunch of unaligned pictures of a common theme. New software can also merge different photographs of the same subject into a single shot by -taking the best part of each one." Bye-bye red eye. There's also Malvar's Big Picture application, which can automatically meld hundreds of photos of some environment -to create a humongous panorama" in which one can delve at the finest of resolutions. He demonstrated this using a photograph of the entire Seattle cityscape, into which he zoomed to discover a pair of worker's green gloves lying atop a skyscraper work site.

The piece de resistance was -Photosynth," recently released on the web, which enables disparate users to piece together -group images that share a context" into a single, three-dimensional image. Imagine hundreds of separate tourist photos of a famous site like St. Peter's Square, assembled into an organized whole by a digital higher power. Users can navigate through the entire environment, piece by piece or as a unified space. It is -augmented reality," says Malvar, -the collision of virtual and real worlds. People can go and visit to see things they've never seen before."
About the Speaker(s): Henrique S. Malvar oversees research in several areas, including Microsoft's Adaptive Systems and Interaction, Communication and Collaboration Systems, Data Management, Exploration and Mining, Databases, Interactive Visual Media, Knowledge Tools, Machine Learning and Applied Statistics, Natural Language Processing, Speech Technology, and Text Mining, Search, and Navigation.
Malvar received a Ph.D. in Electrical Engineering and Computer Science from MIT in 1986. Before coming to Microsoft, he was Vice President of Research and Advanced Technology at PictureTel. Prior to that, he headed the Digital Signal Processing Research Group at Universidade de Bras'lia, Brazil. He was elected a Fellow of the IEEE in 1997, and received the Technical Achievement Award from the IEEE Signal Processing Society in 2002. He holds over 40 patents and over 20 pending patent applications, and has published over 120 articles in journals and conferences.Host(s): School of Engineering, Research Laboratory for Electronics


MIT World -- special events and lectures