TMCnet News
Your virtual avatar has all your best moves(New Scientist Via Thomson Dialog NewsEdge) DISSATISFIED with your virtual self? Until now the only way to represent yourself in a virtual environment, such as a video game or Second Life, has been with cartoonish avatars, controlled by a joystick, keyboard or sensors. Soon you'll be stepping into a video game, looking and moving just as you do in real life, and interacting with your surroundings simply by moving your body. The system, called GrImage a fusion of "grid" and "image" was developed by researchers at the French national research lab INRIA in Grenoble. To make the virtual experience more intuitive and natural, the team had to work out how to capture a person and their movements in three dimensions and then render them in real time. This has been a long-standing challenge in computer graphics. "There's photorealism out there, but it's not in real time: there's real-time motion capture out there, but it's not photorealistic," says Richard Broadbridge of 4D View Solutions, a company spun off by INRIA to bring the technology to market. Most existing methods of motion capture require a person to don high-visibility markers, enabling the relative position and movement of different body parts to be recorded by video camera in 2D and easily rendered. The INRIA team wanted to do away with the markers. Instead, they used between six and 14 video cameras to capture the person from multiple angles and created software that identifies and extracts their silhouettes and combines them to form a rough 3D model. The software quickly adds colour and texture to the model, resulting in a fleshed-out avatar that mimics the subject's movements in real time. The next challenge was getting the avatar to interact convincingly with a virtual environment. Programs that simulate interactions with virtual solids tend to be separate from those that simulate interactions with fluids and soft objects, because they require different mathematics. That limits the variety of objects that can be simulated. GrImage, however, is different, switching mathematics to simulate all types of object. A small version of GrImage was displayed last week at the SIGGRAPH conference in San Diego, California. Visitors could stick their hands in a "recording space" lined with cameras and play with a virtual jack-in-the-box as well as some soft, squishy objects. That impressed John Sibert, a computer scientist at The George Washington University in Washington DC. "When you stick your hand in, the hand on the screen is there immediately," he says. A larger version of GrImage, capable of modelling a full person, is already up and running at the INRIA lab. Besides gaming, the system could also be used for video conferences, says team member Bruno Raffin, allowing people in separate locations to meet in the same virtual space. Copyright 2007 Reed Business Information - UK. All Rights Reserved. |