Field Blending Fix in AviSynth (Script Advice pls)

[luciofulci]

Hi Everyone,

I am a very novice AviSynth user. I can do some basic functions like IVTC but that is about it. I’ve come across a video that I would like to fix. It suffers from field blending and I’m hoping that it’s possible to correct this with AviSynth.

It was recorded from Italian TV at some point in the 80s to VHS and then later digitised, but I have no idea who did this originally so I don’t know what process they were through or at what point the field blending occurred. However it is next to impossible for me to get hold of another source, as for a long time I couldn’t even get this.

I’ve attached a sample video with lots of motion.

I’m spoken to someone who has 100 times more experience with AviSynth than me, but they admin they are not an expert in field blending fixes and suggested I post here as there are some members who are very experienced in this.

This is the script I have tried so far:

LoadPlugin("DGDecode.dll")#from DGIndex
LoadPlugin("masktools2.dll")
LoadPlugin("grunt")
Import("srestore.avsi")
Import("smdegrain.avsi")
Mpeg2Source("test.demuxed.d2v")
#SeparateFields()
Bob()#substitute QTGMC(Preset="Placebo") or whatever
SRestore()

Initially I tried Bob, but the other user suggested it wasn’t a good method and it didn’t make much different, it actually made it look worse. Then I tried QTGMC, however that produced an error from QTGMC “There is no function named “mt_makediff”. I am using VirtualDub2 32 bit and version 3.368 of QTGMC.

Can anyone suggest anything I can do to fix this video as best as possible? As above, I appreciate the best option would be to get a different source but that just isn’t possible for me.

Thanks to anyone who replies.

[ProWo]

The video you have enclosed is progressive and does not require deinterlacing.
Maybe it has been wrongly deinterlaced and recoded before.

[Anonymous344]

Make sure that you are loading the correct .dll: x64 (for 64-bit) or x86 (for 32-bit).

[abolibibelot]

Unfortunately, there's probably nothing that can be done to improve this. Not a specialist, but it would seem like fields from adjacent frames have been consistently blended together, and encoded as progressive, so it's no longer possible to properly process individual fields, because the lossy encoding produced strong artifacting, especially considering that the quality is quite low to begin with.

As for QTGMC (which probably won't help much here but still might be useful later on), it's a PITA to set properly, it requires a number of additional plugins to work, if you open the .avsi file it should indicate which plugins are required for that particular version (search "additional plugins").
I currently have v. 3.358 and it says :

# Additional plugins:
#	NNEDI2, NNEDI, EEDI3, EEDI2, TDeInt - if selected directly or via a source-match preset
#	yadifmod2 - for Preset="Ultra Fast" or if selected directly
#	FFT3DFilter - if selected for noise processing
#	dfttest (1.9.4 or above) - if selected for noise processing
#		For FFT3DFilter & ddftest you also need the FFTW3 library (FFTW.org). On Windows the file needed for both is libfftw3f-3.dll. However, for FFT3DFilter
#		the file needs to be called FFTW3.dll, so you will need two copies and rename one. On Windows put the files in your System32 or SysWow64 folder
#	KNLMeansCL - if selected for noise processing
#	AddGrainC - if NoiseDeint="Generate" selected for noise bypass
#	chikuzen PlanarTools - if you work with YUY2 source (not necessarily required but may improve speed)

[jagabo]

Yeah, you're not going to be able to fix that. About the best you'll be able to do is resize it to half height, then back to full height to blur away the residual comb artifacts (leaving "double exposures").

Mpeg2Source("test.demuxed.d2v") 
Spline36Resize(width, height/2).Spline36Resize(width,height)

It was originally 25p PAL but was captured out-of-phase. It was then resized vertically without any consideration for the interlacing. Find a better source.

[luciofulci]

First of all just wanted to say thanks to everyone who took the time to download the sample and post in the thread - much appreciated.

Richard1485 thanks for the info on QTGMC, whilst it hasn't helped here it is something I want to get working in case I need it in future projects.

Jagabo - that has helped massively! Here is a before and after, might be my eyes but I think it looks a lot better - you might have to click on the images to see the differences better:


[Attachment 54173 - Click to enlarge]


[Attachment 54174 - Click to enlarge]

Still not perfect obviously. And yes a better source would be the best solution, the problem is I am relying on someone who recorded it when it aired back in 1987 ish in Italy, then kept the recording and hangs around forums where this sort of stuff is discussed. I've been looking for a better source for a couple of years now, myself with English speaking collectors and a friend in Italy with Italian speaking collectors - unfortunately we've both come up empty handed. But all considering I am very pleased with the outcome - I owe you a pint Jagabo.

[johnmeyer]

This is somewhat similar to the problem of trying to repair interlaced video that was resized without first deinterlacing. In fact, it may be exactly that. I created a pretty decent fix for this over at doom9.org:

repair bad deinterlacing

I think the same idea might work. It is obvious that in the bad scan lines are from previous fields and that if you can motion estimate those fields back to their original horizontal position, you might get an improvement.

I spent a little time on it, and manage to make it look worse, but I made it look worse in the identical manner, meaning that replacing with adjacent fields from the previous frame did indeed cause the problem.

[Anonymous344]

It was originally 25p PAL but was captured out-of-phase. It was then resized vertically without any consideration for the interlacing.

That makes perfect sense. Every frame was combed, and the vertical resizing baked in the combing and effectively deinterlaced the video (or ineffectively, as the case may be) . No wonder every field has artifacts.

[abolibibelot]

That makes perfect sense. Every frame was combed, and the vertical resizing baked in the combing and effectively deinterlaced the video (or ineffectively, as the case may be) . No wonder every field has artifacts.

Why vertical resizing, since the resolution is standard PAL 720x576 ?
What is surprising is that it seems to have been cropped from 4:3 to 16:9...

It is obvious that in the bad scan lines are from previous fields and that if you can motion estimate those fields back to their original horizontal position, you might get an improvement.

With a lossless capture, or at least high bitrate MPEG, and a reasonably sharp picture, perhaps, but here we're talking about a ~2700kbps MPEG encode of a very blurry source.

I spent a little time on it, and manage to make it look worse, but I made it look worse in the identical manner, meaning that replacing with adjacent fields from the previous frame did indeed cause the problem.

Not sure I understood that part...

[Anonymous344]

Why vertical resizing, since the resolution is standard PAL 720x576 ?
What is surprising is that it seems to have been cropped from 4:3 to 16:9...

Your statement about the cropping is probably the answer to your question.

[luciofulci]

Yes whoever made the disc originally certainly made a pig's ear of it. The audio was also 5.1 for some reason, with the additional 4 channels duplicates of the original two.

Johnmeyer thanks for the link to the other thread, I've had a read through it but I'm afraid it looks to be outside of my skill set.

This mini series has actually been released/broadcast several times in different parts of the world and each version has some scenes exclusive to it - that is why it was important for me to try and fix it or at least make it a little easier to watch.

Thanks again all

[davexnet]

Yes whoever made the disc originally certainly made a pig's ear of it. The audio was also 5.1 for some reason, with the additional 4 channels duplicates of the original two.

Johnmeyer thanks for the link to the other thread, I've had a read through it but I'm afraid it looks to be outside of my skill set.

This mini series has actually been released/broadcast several times in different parts of the world and each version has some scenes exclusive to it - that is why it was important for me to try and fix it or at least make it a little easier to watch.

Thanks again all

What is this show?

[luciofulci]

It's Treasure Island In Outer Space

[davexnet]

It's Treasure Island In Outer Space

I found a 3 hour 4 minute video on Youtube, but it looks to be from a similar source
as yours, with a heavy noise reduction employed.
https://www.youtube.com/watch?v=Asr0hhzOCqI

[abolibibelot]

I found a 3 hour 4 minute video on Youtube, but it looks to be from a similar source as yours, with a heavy noise reduction employed.
https://www.youtube.com/watch?v=Asr0hhzOCqI

Too bad it's also savagely cropped, that guy at 0:10:17 with his gigantic eyebrows would have been so meme-worthy... I can only imagine a montage with the legendary “Dramatic Chipmunk” video...

On a more serious note, there may be a small possibility that the uploader is also the owner of the original recording and still has it somewhere. Problem is, it seems no longer possible to send private messages on YT.