AviSynth or GPU based deinterlacing / IVTC

[topsyturvy]

Hi everyone,

I'll appreciate your feedback on a question that is bugging me. I have 2 types of interlaced mpeg2 content 1) HDV Camcorder 1080/30i family videos and 2) HDTV anime caps also 1080/30i (telecined, orig 24fps)

Since I play these videos through a computer (htpc) using dxva acceleration, my TV detects the connection as "progressive". This means that the videos cannot be played interlaced and must be processed either (a) during playback by the video card (shader based deinterlacing) or (b) pre-processed using avisyth to deinterlace/ivtc and re-encode.

Uptill very recently I was not aware of option (a), and even now I cannot really test it since my video card is a little dated and unable to high-quality deinterlace when playing using DXVA. I've tried option (b) for both types of videos and the results look OK, its a little hard for me to tell since my TV is also rather crappy (only 720p) and is also due for an upgrade when I can spare the cash.

Because of the above constraints I can't test the 2 different methods my self so would like your feedback on which method (1) or (2) above should be better "in principle". Runtime "vector-adaptive deinterlacing" using one of the newer video cards from nvidia or amd, or preprocessing the videos to do, for option (a) smart bob (2x frame rate) + maintain pixel AR, and for option (b) IVTC to 24 fps.

Even with GPU improvements, I find it hard to belive that runtime GPU deinterlacing will be as good as software based smart bob which takes ages even on a modern cpu. Also, the GPU isn't going to IVTC during runtime.

Moreoever my understanding is that using a sufficiently high bitrate with x264 there should be no visible difference in quality since mpeg 4 is much more efficient than mpeg 2. The HDV recordings are about 13GB/hour so this should reduce the file size by over 50%, why waste storage space if there is no real benefit.

What would you do in my stead? I realise that most of us will have 65 inch TV 10-15 years down the line, so if a re-encode looks good today on my current TV it might not turn out to be such a good idea 10 years down the line ...


Thanks and looking forward to your feedback...

[topsyturvy]

No comments .. poor question?

[jagabo]

I doubt any realtime deinterlacer will approach the quality of QTGMC() any time soon. Most of them do little more than a simple bob. On the other hand, any problems introduced by a software deinterlacer will be locked into your deinterlaced footage forever. If the video is important to you I'd keep the original interlaced files. Maybe make a deinterlaced version for day-to-day viewing, distribution, etc.