A lot of time and effort has gone into television production to try to make the fluid motion of video look like film. The differing frame rates of video and film make converting between the two fraught with mathematics: for instance, film runs at 24 frames per second, progressive. This means that there’s one full picture displayed every 24th of a second. American television runs at 30 fps, interlaced. (Actually, it’s 29.94 fps, but who’s counting?) This means that a television displays a full picture every 30th of a second, but it’s made up of two interlaced images every 60th of a second. The television displays every other line of the image, then goes back and fills in the other half. This makes the motion on television more “realistic” and fluid than a film picture.
Anyway, I could go on and on about the ways that film is converted to be shown on television, but that’s not the point of this entry. The interesting thing I want to write about is a technique that no one has bothered to try until now.
In the era before videotape became inexpensive, television programs were saved for future use by using a kinescope, or film recorder. Basically, it was a specially engineered system that, put simply, placed a film camera in front of a video screen and made a film copy of a television program. This was a method of archiving a show as well as selling it to other countries. Because there are different video standards throughout the world, but film had one standard, the film copies were universally accepted.
The major problem with this process is that information is lost. The video signal had more frames per second than the film could handle, so the fluid motion of the video was lost forever. Or so we thought.
A team in the UK working on restoration of some of the original Dr. Who episodes from the early 60s had a simple idea that no one had had before: why not use a computer to convert the 24fps film recordings of old television material to the higher-frame rate interlaced format, thus restoring the look of the original broadcast? They called this process VidFIRE, and have begun using it on DVD releases of old Dr. Who episodes, with amazing success.
VidFIRE was developed to try to reverse this process – or at least to give the illusion that the process has been reversed. Using motion-estimation techniques, pairs of film frames are examined and the paths of moving objects in the frames is calculated. A new frame is estimated, midway in time between the two existing frames, using software that calculates where objects will have moved to 1/50th of a second after the first frame. By taking one video field from the real film frame and the second from the subsequent motion-estimated frame and interlacing them together, pictures with a temporal resolution of 1/50th of a second have been recreated and the ‘video look’ is re-established.
I purchased a DVD of a Dr. Who episode from 1964, and after seeing the shows in their film recorded format, this new VidFIRE version is astounding – clear, sharp, and realistic. It’s like watching the original broadcast rather than a copy of a copy.
This phenomina is why when you buy a tv you want to look for a feature called a 3D digital comb filter, or 3:2 pulldown ratio. Most ED or HDTV’s will have this already and you really shouldn’t buy anything but those, that is to say if you’re a geek like myself.
Oh, how techie.