Deinterlaced source but telecine remains!

[datauser]

Forgive me if I have posted in the wrong section. Anyway I wanted to back up one of my DVD's, it was raining and I needed something to do! After preparing the vobs I found that despite being mostly R2 25fps progressive, it had 66% progressive with 31% interlace and 3% telecine frames. Encoding it as an avi just to get the pattern in AutoGK showed this, FU Wizard confirmed it as mostly native with interlace and HDConverttox interlace test as well.

As I needed to sharpen it up a bit I decided to use HDC as it has advanced avisynth options like film sharp. I did not post there as program is not supported anymore. I set the interlaceHQ, I suppose I could have used the progressive with telecine or auto which does not work according to HDC's guide. I could not find a reasonably quick solution to solve this hybrid dvd and give me better options, but the resulting encoding came out superb as a H264 High 3.1 encoding. I checked its interlacing pattern to find 97% progressive, no interlacing but the 3% telecine remains. Maybe I should have used some pull down or something, but the video looks truly excellent no jagged lines or juddering. :confused: I thought deinterlacing would have removed the telecine as well?

Anyway everything is blended fine so being curious, I also did an FU Wizard encoding which encoded 100% progressive, but the image was a little too soft and grainy for my liking, but good job nonetheless. It used Native and Bob to achieve this. First PAL DVD of mine to be a strange hybrid! My question is how can it still be telecined slightly and play back/look so well that smoothly? Cheers.

Thinking now, I could have used Handbrake with telecine and decomb set, but then again it has no sharpening, just a denoiser!

[huhn]

a 5 sec sample from each type of interlacing place.
it's hard to judge this way.

why not leaving it interlaced any proper hardware deinterlacer can easily deal with 50i there shouldn't be any 3:2 cadence issue like it would be with NTSC.

[datauser]

You read my mind, I was only just thinking of that 5 mins ago! The recommendation is often to leave it interlaced and would not have arising issues it might have had, had it been NTSC. Thanks.

[hello_hello]

Sometimes progressive PAL DVDs can appear interlaced, or partly interlaced, because the fields are out of order (phase shifted). A program analysing them will probably decide they're interlaced, but field matching is all that's really needed (no de-interlacing required). ie

tfm()

It might explain why the analysis is seeing a small percentage as telecined. I'm not sure about that.

I've seen "euro" pulldown used for PAL DVDs now and then. Wikipedia says 24fps is converted to 25fps with this pulldown pattern: 2:2:2:2:2:2:2:2:2:2:2:3.
From memory, you end up with a pattern of 11 clean (progressive) frames and 14 dirty (interlaced) frames. I don't think any program will analyse that and get it right. Euro pulldown can be reversed with field matching, which gives you a progressive output with 1 frame in 25 repeated. The repeated frame can them be decimated. ie:

tfm().tdecimate(cycle=25, cycleR=1)

Another possible "what's going on" check for PAL might be to apply bob or double frame rate de-interlacing. ie:

Yadif(mode=1)

The output would be 50fps progressive. If the video is really interlaced, that'll give you 50 unique progressive frames (hopefully if you de-interlace that's the way you do it). If it's progressive that'll still give you 50fps but I think every second frame will be a duplicate (effectively 25fps with each frame repeated) in which case it'd be an idea to try field matching rather than de-interlacing.

For PAL they're a few possibilities. It can be easier to work out if you're using a GUI that lets you modify the scripts it creates and preview them. I mostly use MeGUI myself. After opening and indexing a set of vob files, choose "Interlaced" in the interlacing section in the script creator and "Yadif with bob" as the de-interlcing method, and it'll add Yadif mode1 to the script. For experimenting with frame matching and decimation if you choose "Film" and "TIVTC" it'll add standard NTSC pulldown removal and decimation to the script. ie

tfm(order=-1).tdecimate()

From there you can delete .tdecimate() to try field matching on it's own, or modify it for Euro pulldown removal if that's what's required, then click the preview button to see the result. :)

[datauser]

I read that interlace flags can be there too despite progressive source dvd. Isn't all this a nightmare for us mere encoding mortals! :eek: I may give it a bash with Bob yadif setting in HDConverttoX with less sharpening(soft sharp)and keep motion filters off, just using super light denoiser as on close viewing of my previous telecine result faces look a bit waxen. FUwizard using that Bob setting did a fine job encoding the source as 100% progressive and removed the telecine with the interlacing, but left it just a little too noisy.

Thankyou for taking the time to give such a lengthy and eclectic explanation.:thanks:

[foxyshadis]

Before doing anything drastic, you should step down to a basic:

Source()
SeparateFields()

And step through the fields one by one in different places, especially wherever your metrics tell you the interlacing changes. Your eyes are still better than any field matchers at being able to tell if field order is correct, what matches, and what the cadence is. If you see backward motion always, use AssumeTFF() or AssumeBFF() before SeparateFields() (try both or use Info() to see which is assumed; by default it's BFF, but some sources modify that). If you sometimes see forward and sometimes backward, you have missing fields due to bad edits; sometimes TFM can compensate, often not. There's no field matcher that can fully handle a floating field order. It's best to manually find the damaged cuts, remove or duplicate a field, and Weave(), or split the video into sections that are individually TFM'd with different orders.

It's also possible that it was converted to PAL from an already telecined NTSC, by dropping random fields. There's no way to fix that, just have to live with the results or find a better source.

[datauser]

You're right I need to figure this out or just live with it. I tried the Bob yadif in HDC and frames doubled to 50fps in the encoding, ran the interlace pattern check on it at the end, still 97% progressive, no telecine, but now 3% interlace. I might give MeGUI a bash as there are some good options or try Handbrake at default telecine and decomb too. I was mystified as to why my frames doubled up using Yadif Bob in HDConverttoX when FU Wizard uses Bob(I assume Yadif 0?) and they remained at source 25fps after a test encoding, but then realized HDC must have used Yadif 1 instead to make 50fps. That's life!:) Thankyou.

[hello_hello]

So even after de-interlacing the whole thing with Yadif mode 1, about 3% still shows combing? Or do you mean 97% of it ended up 50fps with each frame repeated, while 3% de-interlaced to unique frames? Would you care to upload a sample......

Co-incidentally, I came across another PAL variation today. I think it was interlaced 23.976 converted to 25fps using frame blending. I'm not sure exactly what process of butchery was required for that conversion as I haven't quite got my head around the pattern of blurred, combed frames yet, but Yadif(mode=1).SRestore() did the trick for most of it. There's a few small sections though where frames were removed that shouldn't have been (the motion is jerky) so I guess those sections were really just interlaced.....
Back to the drawing board tomorrow. It's the "making of" from the DVD and they tend to be a patchwork mess at times. I might need to upload a sample myself if I can't work it out.