Deinterlacing HD material

[rtjnyoface]

Is deinterlacing HD material the same as SD material? I've recently started capturing OTA HD material and it's 1080i but it seems to be entirely interlaced (using the SeparateFields and AssumeT(B)FF SEEMS to indicate entirely interlaced material).

[thetoof]

If you do have interlaced material... then yes, but slower ;)

Some good to know points though:
- If you use motion compensated deinterlacing with mvtools, using a higher blocksize than SD is recommended for proper motion estimation
- If you use very high quality deinterlacing, be prepared to do a rendering pass of deinterlacing alone, as memory requirements might be too much for additionnal filtering (TGMC runs at .88 fps on a 1920x1080 clip with mem usage at 1450mb and virtual memory at 1500mb single threaded on my 2.4ghz Q6600) As you may know, there's a 2gb memory max per process in XP (which you can remove in some way, but I don't recall how) that's always good to keep in mind.

There could be other things, but this is all I can think of for now.

[knutinh]

Did anyone compare ATIs "vector adaptive deinterlacing" with avisynth tools?
http://anandtech.com/video/showdoc.aspx?i=3734&p=4

ATIs advantage: having high-power dsp-like/fixed-function hardware available "for free".

Avisynths advantage: not caring about realtime, large group of dedicated tinkerers, easier to develop for "expert users" than "housewhifes/husbonds".

-k

[aegisofrime]

Did anyone compare ATIs "vector adaptive deinterlacing" with avisynth tools?
http://anandtech.com/video/showdoc.aspx?i=3734&p=4

ATIs advantage: having high-power dsp-like/fixed-function hardware available "for free".

Avisynths advantage: not caring about realtime, large group of dedicated tinkerers, easier to develop for "expert users" than "housewhifes/husbonds".

-k

Unfortunately AFAIK ATI does not allow third-party access to the deinterlacing functions, unlike nVidia. That's why you have DGNVTools, but nothing like that for ATI.