PAL DVD video looks like it has ghosting (Trying to get rid of or reduce)

[darkdream787]

I bought a PAL DVD set because it wasnt released in NTSC or in my country so I cant watch it in my DVD player untill I encode it. Ive only ever tried to rip and encode one other PAL source and it did this ghosting thing too. I ended up just letting it go because I couldnt get rid of or reduce it. It seems all PAL DVD sets do this, Im guessing because the footage was sped up to 25fps.

Ive tried QTGMC(preset="slow") but it really isnt helping me a noticeable amount.

Is this just me or is that PAL and Im gonna have to deal with it?

Heres a 1 minute video sample.

http://www.sendspace.com/file/n2nqba

[sanlyn]

That's not ghosting. It's a telecined video. Play it on a player that honors pulldown flags and/or has deinterlace turned on. Deinterlacing doesn't get rid of it. Use:

TFM().TDecimate()

[darkdream787]

That's not ghosting. It's a telecined video. Play it on a player that honors pulldown flags and/or has deinterlace turned on. Deinterlacing doesn't get rid of it. Use:

TFM().TDecimate()

So should I put that before or after the Yadif(order=1) deinterlacing? Does it even matter when thats called in the script?
Ive only used that once before because of a suggestion but never asked what it was for or did.

[sanlyn]

If telecine is removed, your vid will be progressive. Removing telecine comes first.

[manono]

It's field-blended from an NTSC source. This does a decent job of unblending it:

QTGMC(preset="slow")
Srestore()

You don't really have to use the slow preset, unless you really want to. One of the faster ones should do almost as well (besides being faster to encode).

If telecine is removed, your vid will be progressive. Removing telecine comes first.

It's PAL, sanlyn.

[sanlyn]

Nope. Just tried it myself. On closer look, your vid is field blended. You've got a big problema. Will have to try it another way.

[darkdream787]

If telecine is removed, your vid will be progressive. Removing telecine comes first.

So by using the filter you suggested I wont even need to deinterlace with Yadif anymore because its turned to progressive. Is that correct?

[darkdream787]

It's field-blended from an NTSC source. This does a decent job of unblending it:

QTGMC(preset="slow")
Srestore()

You don't really have to use the slow preset, unless you really want to. One of the faster ones should do almost as well (besides being faster to encode)

I was trying that exact thing before even asking and I didnt notice it being much of a help. You sure thats good enough?

[sanlyn]

Yep, manono, it's PAL. You beat me to it with the script while I was giving it a try. I do see a couple of blended frames remaining. Shucks, there must be break in the blend sequence. Frame 51, and some more after frame 90.

Darkdream787, it's not ghosting. It's blending.

[manono]

I was trying that exact thing before even asking and I didnt notice it being much of a help. You sure thats good enough?

Yes. For testing use:

Yadif(Mode=1)
Srestore()

because it's way faster. And don't move all around in the time line. Try this. Open the sample alone with that script and encode it with something. Then go through the resulting AVI or MKV or whatever you make. Or just work your way through the sample from the beginning. Just don't jump all around. If you skip ahead to somewhere, stop short of it and then advance on it. Srestore gets confused by people moving back and forth and expecting the frame you stop on to be 'good'.

SRestore isn't perfect, especially at scene changes. But it's pretty decent and the chances are you won't see any problems when playing it as opposed to studying it frame by frame. And there are work-arounds available for the scene changes.

[sanlyn]

I tried this and got 23.971 FPS:

AssumeBFF()
yadif(order=0,mode=1)
TDecimate(mode=2,rate=23.97)

I get one blended frame (scene change) and a few with slight blending remnants. Plays fine in any player with deinterlace turned on. Some kind of crazy blending/deinterlace sequence, I guess.

[sanlyn]

Got the attached result with this code. Ended up with one blend, a few vague blend artifacts that run too fast too see, and also stopped most of that frame hopping.

MPEG2Source("[path to video\dgindex.demuxed.d2v")
yadif(order=0,mode=1)
TDecimate(mode=2,rate=23.97)
Stab()
Crop(12,2,-12,-2).AddBorders(12,2,12,2) #<- clean up borders after Stab()

[darkdream787]

Got the attached result with this code. Ended up with one blend, a few vague blend artifacts that run too fast too see, and also stopped most of that frame hopping.

MPEG2Source("[path to video\dgindex.demuxed.d2v")
yadif(order=0,mode=1)
TDecimate(mode=2,rate=23.97)
Stab()
Crop(12,2,-12,-2).AddBorders(12,2,12,2) #<- clean up borders after Stab()

Unfortunatley that didnt turn out too great or has more side effects than Im willing to accept. The colors look like their bleeding in cube shapes over the outer black lines (pause at 31-32 seconds and look in the girls red hair. Thats not the only place you see it) and the interlace lines look like their still there, especially when they talk. It did seem to help the blending and ghosting though.

Im trying 8 different scripts and methods. Ill post the results of my favorite one and the script as well when I get it.

Thanks for the info everyone.

[sanlyn]

True, it will take some study to come up with a definite pattern of frame blending. You'll find a great many posts on the subject, here an on doom9, over the past 10 years. More than loikely, this video was originally progressive, then was telecined, and some idiot field-blend deinterlaced instead of removing the telecine, then probably discarded 50% of the original fields. So there's lots of material on the subject.

However...blending and ghosting are not the same thing: de-blenders and decimation plugins won't do anything about ghosting, nor will they prevent chroma bleed.

[darkdream787]

Well srestore() helped but it was making the video jumpy when the entire screen moved so I ended up not liking it. Instead of a smooth slide up, down, left, or right when the entire screen moved it had slight jumps. I couldnt stand that. Didnt matter what combination or how I used it to lessen that, it always did it to the video. Thats not the first time I decided I didnt like srestore() for that reason.

So far the best one to me was just leaving it alone and sticking with Yadif.

[manono]

Well srestore() helped but it was making the video jumpy when the entire screen moved so I ended up not liking it.

It's that simple animation responsible for it, not SRestore. Much of it (the sample anyway) is drawn at 8fps or even less so you get three frames (or more) the same before it changes and that's what makes it play jerky. What you're saying is you prefer the blurry but smoother deinterlaced appearance rather than the way it's meant to be viewed.

There weren't any real panning shots (which is what you described as when the entire screen moved, but if even they play jerky it's most likely because they weren't drawn at 24fps, but 12 or even 8fps. Better quality animations create 24fps pans.

To each his own.

[darkdream787]

Well srestore() helped but it was making the video jumpy when the entire screen moved so I ended up not liking it.

It's that simple animation responsible for it, not SRestore. Much of it (the sample anyway) is drawn at 8fps or even less so you get three frames (or more) the same before it changes and that's what makes it play jerky. What you're saying is you prefer the blurry but smoother deinterlaced appearance rather than the way it's meant to be viewed.

There weren't any real panning shots (which is what you described as when the entire screen moved, but if even they play jerky it's most likely because they weren't drawn at 24fps, but 12 or even 8fps. Better quality animations create 24fps pans.

To each his own.

Actually I did it like this and this was the second best one out of all the different ways I tried it.

Yadif(order=1)
QTGMC(preset="slow")
SRestore()


I took out the yadif and only used these 2 and it seems to play better so far according to preview, Ill let you know when its done encoding if its better.

QTGMC(preset="slow")
SRestore()

[manono]

It's field-blended from an NTSC source. This does a decent job of unblending it:

QTGMC(preset="slow")
Srestore()

That was my first post in this thread. And then you have the nerve to say, "Well srestore() helped but it was making the video jumpy when the entire screen moved so I ended up not liking it.", implying Srestore was at fault when in reality you didn't even use that script? Instead you used a much worse one involving 2 different bobbers with the result that "when the entire screen moved it had slight jumps"?

Ill let you know when its done encoding if its better.

Me, I couldn't care less, not when you said the provided script didn't work and it turns out you didn't even use it.

[darkdream787]

It's field-blended from an NTSC source. This does a decent job of unblending it:

QTGMC(preset="slow")
Srestore()

That was my first post in this thread. And then you have the nerve to say, "Well srestore() helped but it was making the video jumpy when the entire screen moved so I ended up not liking it.", implying Srestore was at fault when in reality you didn't even use that script? Instead you used a much worse one involving 2 different bobbers with the result that "when the entire screen moved it had slight jumps"?

Ill let you know when its done encoding if its better.

Me, I couldn't care less, not when you said the provided script didn't work and it turns out you didn't even use it.

I did use the script mr cocky!! along with several others and changing the orders they were used in the script as well untill I had about 12 different vids to compare. I just never heard anyone say take out the Yadif as well. From the way I got it, all you were saying was add it in where it should go because it does pretty well, not erase anything from my script as well so I put what you said right in my script after the deinterlacer exactly as you typed it. I thought those were supposed to fix the blending (I wasnt asking about fixing deinterlacing), I didnt know it would deinterlace by itself as well as yadif on that source even though QTGMC is a deinterlacer. I havent exactly used or needed to use QTGMC but one other time so you couldnt expect me to know everything about it and its capabilities. IVTC and or Yadif have always worked for me when it came to deinterlacing my anime DVD's. I cant say I knew using QTGMC with another deinterlacer would be a problem.

Srestore() still makes the video jumpy but when I took Yadif out, it jumps a bit less to where you can still see it but its not as visible as before. So like it or not the results are still the same except now its less jumpy and a little more smooth moving.

The only real bonus I got out of taking Yadif out is the picture seems to look a little more sharp or clear so /softness/bluriness is a very tiny bit reduced and the jumping lessened to where its plenty more acceptable. I like it more now than I did before.


Next time, instead of being the impatient professional then getting angry or annoyed over something petty like that. Try a little patience with the asker and make sure they fully understand how their expected to use it might help some. It turns out it was just a small mistake on my part from not realizing yadif was supposed to come out when you typed that script and not fully knowing about QTGMC from lack of exp with using it. No need to get all defensive and claim you couldnt care less because I didnt use it the way you meant it to be used on accident.

Your a good help and I know your name well from help for several years but relax some lol.

Thanks!! I got the results that I for the most part was looking for.

[manono]

Srestore() still makes the video jumpy

No it doesn't, not any more jumpy than the video is to begin with. If you think it does then provide an unprocessed sample of a section you think it makes jumpy and the complete script you used. You seem to be saying it removes unique and necessary frames and I don't believe that's the case here. There's lots of light residual blending, yes, because this thing is a mess, but I don't believe the sample you provided plays jumpy after using Srestore properly.

So, if one bobber is good then two must be better? Was that the thinking? Yes, I know you're inexperienced, but when making claims about something, or explaining the results of something, it's pretty basic that one posts the script used so the problem can be replicated. I didn't get mad because you were inexperienced but because you were making a claim about jumpy video when you used a completely ridiculous script to create this jumpiness without even saying what you did, leaving us to believe you had used one of the suggested scripts. Nowhere in this thread did anyone suggest using 2 bobbers before SRestore.

I thought those were supposed to fix the blending (I wasnt asking about fixing deinterlacing)

Unblenders (because they use bobbers), by definition get rid of interlacing. Frames become interlaced because they are made up of 2 fields from different points in time. Bobbers take each field and make a full frame from it, in the process doubling both the frame count and the frame rate. The fields themselves aren't interlaced so neither are the frames made from them. Then an unblender such as SRestore chooses one of those frames to display and also does any decimation necessary. With luck it chooses an unblended frame to display.

But if both fields of a frame are blended (and many of yours have at least light blending on both fields), then there's no way for an unblender to do its job right. It should choose the field/frame that's more lightly blended. And a lot of times a blended frame at a scene change is chosen because the 2 fields before and/or after the scene change might be blended.