3D video production now possible with a normal film camera… Goodbye Mo-Cap!
Researchers from Texas A&M University have come up with a solution that can handle realistic 3D animation and extract 3D models from a normal 2D video taken with an ordinary film camera.
The scientists, Xiaolin Wei and Jinxiang Chai, have developed a software system that can analyze a 2D video, track the motion of the subject from the video and use the data on a similar 3D model, after creating the model itself. The result is, as we can imagine, nearly as good as a Mo-Cap (motion capture), for the motion is being extracted out of a real actor.
The scholars intend to present their research at SIGGRAPH in Los Angeles this month; the entire paper can be found here [~13MB pdf]. Their algorithm operates using Newtonian physics, and while fed with a video footage with consisting distinct motion keyframes, it parses the motion with references to the keyframes. The advantages of this system is that it removes the need for those precious motion capture camera arrays and reduces the workload on the rendering farms’ workstation computers. The software can work out normal walking to jumping, and even complex motions such as a weight-lifting, gymnastics etc. Watch the video clip for more details.