The ability to digitally transpose a human being's facial features on to the face of another person is no longer an expensive proposition. Kevin Dale, a graduate student at Harvard University, has claimed the creation of a technique that allows technicians to do just that, with minimum investment.
The act of transposition required, originally, expensive software and hardware and was a complex process calling for considerable expertise and manual intervention. This made the technology something of a pipedream for most low budget film makers; who would otherwise have used the technology to portray twins or characters with near-identical facial features and expressions in their films.
Dale and his colleagues presented a research paper - Video Face Replacement - in the journal Transactions on Graphics, in which they pointed out that the new technology could minimize investment and manual expertise required.
According to their research, all that is required is a camera and a minimum of manual intervention to replace facial expressions. The method, in brief, requires the creation of 3D models of both the target and source faces. This is done by using a 3D multi-linear tracking camera. The captured expressions are morphed on to the target's face and the videos are aligned and blended using a novel mesh-centric gradient domain-blending technique that avoids flickering while the video is played.
The process allows for a ten-second video to be created in 20 minutes; on a normal desktop computer. In addition, Dale suggests that it could also be possible to combine the facial expressions of an actor from different shots, for the best possible effect.
What that means is, theoretically, the ability to piece together a face by taking different components - eyes, eyebrows, cheek structure, lips, etc. - from different people.
While the new technology may be a little less than perfect for use in high-end professional film making, it could become a handy took for amateur and low-end budget film-makers.