- capture a digital version of the video signal from your VCR
- encode that to MPEG2 dvd
- transcode the lossy MPEG2 to something a video codec that can go in a .ogg container. edit (apparently just remux, not xcode, so no quality loss at this step after all).
Because of the limited quality of the source, you're probably not losing much quality from having a lossy intermediate codec. Obviously it would be better to go directly from capture -> final, with only lossless steps in between (e.g. capture card in a computer, capture to a lossless file. Or encode in realtime to the final format.)
EDIT: turns out you're just remuxing the dvd-video into an ogg container, not transcoding. I didn't think ogm could hold mpeg2, but I guess it can, and for some reason you're using some software that defaults to muxing into an ogg container.
Unless you're archiving in a lossless format (50x the filesize of visually-indistinguishable lossy), you might want to apply some filtering to clean up the video from the VHS tapes before encoding. Depending on the condition of the tapes, they might have significant chroma noise that a denoise filter could help with. This will require some playing around.
edit: if you're going to stick with a capture device that doesn't give access to the digital video except after it encodes to MPEG2 (and makes a trivial menu to make a complete DVD-VIDEO), xcoding the MPEG2 to lossless would only be useful if you wanted to apply some filtering. Never xcode a lossy codec to lossless. You'll increase the filesize by a LOT. The lossy encode is the most compact representation of that exact decoded bitstream, so just save it instead of a lossless representation of decoding the lossy encode.
So, if you can get deinterlace/denoise filtering to work well enough "on-the-fly" in realtime, then you can use dvd images straight from the Magnavox as your archival masters. In 10 or 20 years, there might be smarter filters that will do a better job, so it's best to keep the source, as long as it's in a state that's usable in realtime.
If you can't get the necessary filtering to run well on-the-fly on at least your library computers, then you should probably deinterlace and maybe even denoise now, and make either lossless h.264, FFv1, or something lossy. mcdeint is slow, but does a good job. (don't just use yadif for master copies.) You'd need to decide whether to archive at 720x480p60, or throw away half the temporal resolution and deinterlace 720x480i60 -> 720x480p30. The decision is somewhat content-dependent, but if you're using a codec with inter-prediction (P and B frames), like h.264, then higher fps means frames are more similar to each other, so it takes much less than twice as much space to compress twice the FPS at the same quality.
http://linux.goeszen.com/digitising-vhs-tapes-on-linux.html has some stuff to say about it.
http://ubuntuforums.org/showthread.php?t=2091899 suggests hqdn3d as a denoise filter. That sounds right; I would have tried that one first over any of ffmpeg's other postprocessing filters, if I was trying to do this myself.
end of edit
