The ability to quickly and reliably transmit 3-D images over the Internet could make everything from shopping for real estate to playing video games a lot more “real.” Imagine clothing companies transmitting 3-D body scans of their customers to a factory where tailored goods are made and shipped directly to the customer. Real estate firms could offer sophisticated virtual tours of homes on the Web, where a potential buyer could “walk” into a house, move furniture around and even tear down walls.
But for the average Web user with a 56K modem, downloading complex 3-D images off the Internet remains cumbersome, with long delays and unexpected crashes. To make 3-D virtual objects more widely available on the Internet, a team of researchers from Lucent Technologies’ Bell Labs and Caltech have demonstrated a way of compressing “digital geometry” that allows 3-D data to be downloaded much faster and more efficiently than is possible with current standards.
Compression of video and audio, squeezing digital data into as few bits as possible, has become commonplace on the Web. But the same compression tricks that work so well for audio and video don’t work nearly as well for geometry, which involves huge and complex data sets. “Geometry is already appearing in many different places on the Internet right now and will eventually become available on a broad basis,” says Peter Schrder, one of the researchers involved in the project and a computer scientist at Caltech. As the demand increases, so will the need for sophisticated tools for transmitting the information over the Net.
Schrder and Bell Labs’ Wim Sweldens (one of the TR100) tackled the problem by designing a mathematical technique using “wavelets”-a fancy way of processing the millions or billions of triangles used to describe a three-dimensional shape. MPEG-4, the latest standard on the market for deploying multimedia content over the Web, sends all the information in one shot. In contrast, the new method transmits information progressively. For instance, when downloading a 3-D representation of a person’s head, the most important information, like the rough shape of the head, arrives first, followed by the finer details such as the shape of the eyes and nose and the texture of the hair. Tests have shown that this method takes one-twelfth the time of MPEG-4.
“They seem to have finally crossed the threshold for delivering complex geometry over the Net,” says Tony DeRose, senior scientist at Pixar Animation Studios in Richmond, Calif.
While the Bell Labs team is looking into the possibilities of commercializing their compression technique, its members declined to comment on when it will be made available. “The technology itself could be deployed in short order,” says Schrder. “We have working code, after all.”
Be the leader your company needs. Implement ethical AI.
Join us at EmTech Digital 2019.