Compressor's "best" deinterlacing

Every video that is projected, or played on a LCD, is deinterlaced. From a quality standpoint, the producer of the video should do the interlacing rather than leave it to some unknown downstream device to do. The downstream device needs to do it in real time, generally with small processing power, and the producer, hurried as he might feel, has more time than that.
Compressor's "best", motion-compensated deinterlacing takes 45 minutes per 1 minute of DV-PAL video to do on my so-so Mac Pro. I can live with that if it is a really good job of deinterlacing. It is noticeably better than Compressor's "better" motion-adaptive deinterlacing, but there are still little jitters. How does Compressor's "best" compare with the best available deinterlacing, assuming processing time doesn't matter?
Thanks,
dcouzin

----------------------------
Note added Feb 20, 2009. There was a big hole in my thinking above. The downstream display device, assuming its refresh rate is 50 Hz, will in effect deinterlace the 50i video to 50p. In theory, a 50p deinterlace of a 50i video can be visually better than a 25p deinterlace of the same video. So, unless Compressor is also making a 50p deinterlace, the display device's real time deinterlace might well be visually better. Based on filesize evidence, Compressor is making a 25p deinterlace. It's a pity, since deinterlacing algorithms can just as easily produce a 50p as a 25p deinterlace of a 50i video.

Dennis Couzin
Berlin, Germany

Hot knowing where you final output is going..... As general speaking I would say this.

Compressor does a good job at de-interlacing. But in that quality tab there are more factors to consider. You don't always want best everything IMHO.

I have a bunch of custom presets in compressor, the best of which i call strypes 7.7 . The 7.7 refers to max bit rate which has a larger effect on your final out put than some of the other quality choices.

Getting good de-interlacing is like working on an assembly line. The line starts on the shoot and ends at final output.

How was it shot 60i or 24p? How was it captured? What codec was used on the TL sequence? How was it exported? Then comes compressor quality and filter settings settings.

Generally, the final out put or end user result has less to do with the processing power of say a dvd player, than does the pipeline the video came from.

""" What you do with what you have, is more important than what you could do, with what you don't have."

> > > Knowledge + Action = Wisdom - J. Corbett 1992 """"

Thanks J Corbett. The video was shot in DV-PAL, so it is 50i. Interlacing is even uglier at 50 fields/sec than at 60 fields/sec, so I think a good job of deinterlacing can help this work. I was planning to output a 25P (or is it 50P?) QT file from the FCP edit using Compressor. This QT file will serve as a master for later output. I haven't decided whether to leave it DV, or make it 8-bit uncompressed, or even 10-bit uncompressed. Whichever kind of QT file, doing a good job of deinterlacing won't make it any bigger than doing a poor job of deinterlacing, so why not do a good job?

A stamped (not burned) DVD will be one output format. Since the work is just 79 minutes long, I could push the 9.8 Mbps bit rate limit and still fit a single sided DVD. [I posted a question about bit rate limits for stamped vs. burned DVDs in the DVD Pro forum.] Sending a small hard drive with the QT file to theaters is another output format. If a theater has QT on their computer hooked to good video projector, then their audience should get to see something better than a DVD. (If 10-bit uncompressed then it must be a FW800 or eSATA hard drive or the theater must move the QT to their system for playing).

Maxing out all quality settings doesn't make sense when they compete for bits, but deinterlacing doesn't. What I'm unsure of is my strategy of doing just deinterlacing in the first step (making a QT file). Maybe Compressor works better (smarter) when it does everything in one step (e.g. making the progressive mpeg2 from the FCP edit) or maybe a sequence of steps is OK but deinterlacing shouldn't be in the first step.

Dennis Couzin
Berlin, Germany

i suggest you download the Graeme Nattress de-interlacer.
it's much faster than compressor, and it does a better job than FCP's built-in one.

cost is about 100$ for that plus other plugins,
and there are demo versions.

i was at a movie hose in LA a while back, and was a bit surprised to see TV-type ads playing there,
video or data projected, of course.
what made them VERY telvision was the fact that they were NOT de-interlaced, (Or were 60p, perhaps??)

so given that maybe not all projectors do a de-interlace,
you are right - it's best to do it yourself!


nick

Thanks Nick Meyers, for the Graeme Nattress tip, and your confirmation.

What I meant by "every video that is projected, or played on a LCD, is deinterlaced" is that these devices would never output what a CRT outputs, with half the lines black at each instant. They could, but it would result in half-brightness projection or display, an unacceptable waste of light. So they're putting something into every line at each instant. It could be extremely crude deinterlacing, like showing the unmodified even and odd fields at once, but it's deinterlacing nonetheless.

Dennis Couzin
Berlin, Germany