intelaced encoding for standalone DVD player

[oayz]

What is the proper procedure to encode interlaced material to play on Divx/Xvid stand alone DVD players?

My TV has deintelacer which much better then any software one and I would prefer to keep interlaced structure as in original MPEG2. Should I check "interlacing" option in the encoder?

Thanks!

[DivXWill]

yes, you can do interlaced encoding. go to the video tab, under "Interlacing", select "preserve interlace", it's that simple :)

[manono]

Hi-

Unless your TV set has either a Genesis/Faroudja or a Silicon Image chipset for deinterlacing, I might give you an argument about what you have being better than any software deinterlacer. There are some very good AviSynth deinterlacers now. Have you already tried TDeint and/or (Leak)KernelDeint? neuron2 is working on an AviSynth deinterlacer that emulates what the Faroudja DCDi does.

Also, you have to be very careful about resizing when encoding interlaced. But maybe you already know that. Maybe you also know that encoding interlaced material requires a much higher bitrate for the same quality as progressive material. I've read estimates of up to 50% higher bitrate needed. Anyway, I'd be curious about how it works out for you. Please let us know.

[oayz]

Hi manono. Thanks for reply.

I do have deinterlacer in my Pioneer HD533 comparable to Faroudja. It's designed by Aopen (Israeli company recently aquired by Intel).

No I haven't tried software deinterlacers you've mentioned. I'm very intrested in this develompment and will definetly find them and try out. Very intrested in neuron2's work - not many people know what does Faroudja actually do.

In regards of bit rate, are you suggesting that encoding deinterlaced material vs. original intelaced saves bit rate? This is counter intuitive. How would 2 full frames (480 or 576 lines) get compressed better then 2 fileds (240 or 288 lines)? Please elaborate.

Thanks for stating again this golden rule - NO RESIZING on interlaced material. Too many times I have to tell it to other people.

From 2 experiments I've done I haven't seen much difference in quality between "preserve interlaced" on and off. Though I have to admit my material was wierd PAL to NTSC convertion. No 3:2 pull-down :-( Bit rate was average - 1500

[manono]

Hi-

AVIs are always stored as full frames, so all that combing/feathering requires a much higher bitrate to reproduce than if it had been deinterlaced before encoding.

How would 2 full frames (480 or 576 lines) get compressed better then 2 fileds (240 or 288 lines)? Please elaborate.

I may be misunderstanding what you just said, but I wasn't suggesting that you use a doublerate deinterlacer (although plenty of people do that), converting 29.97fps to 59.94fps, or 25fps to 50fps, thereby converting each field into a full frame. I meant for you to try a good standard deinterlacer, keeping the original framerate. So, the 2 original fields would be deinterlaced into 1 frame

From 2 experiments I've done I haven't seen much difference in quality between "preserve interlaced" on and off.

I didn't quite understand that one. Are you saying that you tried encoding it deinterlaced, and then again keeping it interlaced? First, don't use the deinterlacer built into DivX. You'll get much better results using a good AviSynth deinterlacer. Second, try encoding it deinterlaced and then again without deinterlacing it. Do both with a constant quant 2 (for example) and compare the final file sizes. Another way is to run 2 passes of each for a fixed final file size or bitrate, and then compare the average quants in something like DRFAnalyzer (http://www.geocities.com/analyzerDRF/).

[oayz]

This is interesting. I completly unfamiliar with how encoders actually handle video. You're right - combing two interlaced files and then trying to encode such frame makes no sense - waste of bandwidth. So you saying that right way to do is to deinterlace each field to full frame and then drop every other one. I agree that compression and resolution will be good but motion would be jerky since we effectively reduce refresh rate. From this perspective TV deinterlecing has superior quality.

Do you have any idea what does "preserve interlacing" do? I was under impression it compresses each field independently similar to what MPEG2 does. Of cause bitrate wise it may not as effective as your suggestion but at least it's better then combing and doesn't have motion jerkiness.

One other question. If I use your suggestion and end up 30fps progressive Divx how would it be played on Divx enabled player? It would split it in 2 fields and play interlaced, wouldn't it?

Thanks for you help!

Added:

I've took a quick look on TDeint and KernelDeint. None seems to get close to what *faroudja* is doing. The best I've seen so far is TomsMoComp but even it is not as good though some people call it "motion compensated" (it isn't). Could you point where I can find ann info on this new Neuron2 development? Is it Donald Graft?

[manono]

Hi-

So you saying that right way to do is to deinterlace each field to full frame and then drop every other one.

I don't know what the right way to do it is. Interlaced material isn't my thing. I work with movies usually. That's why I said that I'd be curious how doing some interlaced encoding works out for you. But no, the last thing I'd do would be to use a smart bobber (doublerate deinterlacer) and then drop every other frame. I'd just deinterlace the original 29.97fps. TDeint is good, and it doesn't claim to be similar to what Faroudja deinterlacing does. So it will give you jaggy diagonal lines.

If I use your suggestion and end up 30fps progressive Divx how would it be played on Divx enabled player? It would split it in 2 fields and play interlaced, wouldn't it?

To your HDTV? I assume that each full frame would get played 2 times. Mine's hooked up with a DVI cable, and is upscaling to 720p, so the player is outputting the original AVI frames (Faroudja deinterlaced if necessary), I think. If you set yours to output 480i, maybe it'll send fields to the TV set. I don't know for sure. I could be wrong, but I would think that "Keeping Interlaced" would be more for standard interlaced TV sets where each field would get played. I've never experimented with interlaced encoding for AVI, and don't intend to. I'm sure there are others around here that know more about the subject than I.

Could you point where I can find any info on this new Neuron2 development? Is it Donald Graft?

Yes, neuron2=Donald Graft. The filter is available as a VDub plugin:

http://neuron2.net/smart/smart.html

It's an updated version of his Smart Deinterlacer Plugin for Virtual Dub. He's still working on the AviSynth version. Here's a discussion of it (most of which is off topic):

http://neuron2.net/board/viewtopic.php?t=680