Auto IVTC of CAV LDs based on vertical interval data?

In theory it should be possible to perform an inverse telecine with the same accuracy as an LD player's own Pause function using the "white flag" or the frame numbers from the Philips codes.

However, these flags weren't always accurate, leading to jumpy still frames. (Here is a working link to the text file quoted.)

Laserman said:

Now to do this on Laserdisc, they player has to know which two adjacent fields actually make up the frame (othewise you might get one field from one frame of film, and the other field from a diferent frame - not good). So how does an analogue system cope with this? Easy, encode the required information in the VBI (the vertical blank interval). When making the disc, you store the info in the VBI, its often referred to as a 'white flag'. When you hit the pause button on a CAV disc, it reads the flag, and the laser assembly actually does a one track reversal (i.e. 2 fields) and can then redisplay the current frame. It is set in the VBI area outside of picture info, or CC info (its at line 11 or 274 depending on the field). If you get it wrong, the pause feature will have a 'jiggling' frame for 40% of the frames! You can see this on some discs. Sometimes just the 'picture number' is used instead, which is also encoded into the VIL.

"About Laserdiscs"

I have a setup that lets me capture the relevant VBI lines, but it will be quite a while before I have access to my LD player again and I'm not sure whether it's actually worth pursuing. In theory it could help for discs where the pattern breaks often, but those are also probably the releases that are most likely to have the wrong data.

CLV disks don't have white flags - only the LD-V8000 even supported them on CLV, so there wasn't a point to putting it on any movie disks.  I've had good luck so far using ffmpeg's pulldown detection, which supports regular deinterlacing as a fallback.

IIRC they were planning to use the time compressed Star Wars master for the CAV version but they couldn't get the white flags right on it.  So the cadence is probably all wonky anyway.  The JSC is also known for being off, sadly.

The VBI data itself is pretty easy to work with - I have basic but imperfect sample code for the frame #s, and recognizing the white flag is trivial... getting it into a regular workflow is the problem.