Best deinterlacer in 2023 for single-rate deinterlacing?

[orion44]

Which of these deinterlacers produces the best quality in single-rate mode in 2023?

[Sharc]

Depends on the source.

[Selur]

+1 for option 'Sharc', also iirc doesn't https://github.com/realfinder/AVS-Stuff/blob/master/avs%202.6%20and%20up/QTGMC.avsi support using BWDIF and BWDIF+NNEDI3 ?

[Bwaak]

^ This is why I am having a hard time comparing QTGMC with "proper" deinterlacers. QTGMC uses different scripts and libraries depending on the settings, it is an umbrella script.

It is like comparing CD Audio with DVD Audio - the latter supports a whole bunch of configurations including the same one as CD Audio.

[Quint]

And, as said, it depends on the source. Plus your expectations. Plus how you define quality. Plus your requirements as time, hardware...

[lordsmurf]

^ This is why I am having a hard time comparing QTGMC with "proper" deinterlacers.

What does that statement even mean?

QTGMC uses different scripts and libraries depending on the settings, it is an umbrella script.

Open the script file. Read it. It's verbose.

[sophisticles]

As should be obvious to anyone that actually looks at QTGMC, it is not a deinterlacer, it's a script, a front-end that facilitates the processing of video by actual deinterlacers, like yadif, for instance.

It's like asking what's a better encoder, x265 or Handbrake?

[Skiller]

As should be obvious to anyone that actually looks at QTGMC, it is not a deinterlacer, it's a script, a front-end that facilitates the processing of video by actual deinterlacers, like yadif, for instance.

It is not a traditional deinterlacer, yes (more on that later), but you miss the principle of QTGMC's core. It has a completely unique approach, not comparable to any other deinterlacer (Yadif, etc.).

It basically works like this: first it does a dumb bob-deinterlace (with or without more intelligent spatial interpolation such as NNEDI), then temporally blurs the differences in bob-shimmer away and uses motion compensation to detect where to do that and where not to avoid a motion blur effect. That was the original concept written by Didée and named TempGaussMC. It's essentially a bob-shimmer remover, and there is nothing to my knowledge that works even remotely similar. Just because it's not a stand-alone DLL doesn't mean it's not a deinterlacer, imo.

[lollo]

As should be obvious to anyone that actually looks at QTGMC, it is not a deinterlacer, it's a script, a front-end that facilitates the processing of video by actual deinterlacers, like yadif, for instance.

The basic task of a deinterlacer is to reconstruct the full frame from a field having half of the vertical size through line interpolation. In that sense a simple Bob() command is sufficient.

All the deinterlacers, while doing that, try to improve the final output, eliminating the "bobbing", reducing the artifacts, taking care of the edges, elaborating the motion between the fields and so on.

This is true for all options mentioned, and obviously for QTGMC as well, which is a deinterlacer. Or if you stick to the Bob "definition", none of them is.

[Skiller]

The basic task of a deinterlacer is to reconstruct the full frame from a field having half of the vertical size through line interpolation.

Yep, and this is exactly what QTGMC's core does not. That's why (at default settings) it's not a traditional deinterlacer, it's a bob-shimmer remover. It does not just inject new interpolated lines, it alters everything. Is that bad? Most of the time for most people not at all.
Back in the day when QTGMC was still TempGaussMC, for me it was a big drawback because it meant I couldn't go back to "untouched" interlaced fields. The video also looked much different due to the inherent denoising of the bob-shimmer removal.

But with the lossless options and noise retention options added in QTGMC, we're basically at a point where QTGMC's output can be configured to be like that of a traditional deinterlacer, and an excellent one at that.

[lollo]

Yes to all. The “basic” operation is almost useless, and no tools except Bob() does only that.

Lossless and reduction of noise removal is a must for me using QTGMC, even because I prendere to use a specific denoiser after deinterlacing.

I remember the old times of Temporal Gauss Motion Compensated, it was a “game changer” at that time. Master Didèe gave us many wonderful tools…

[sophisticles]

For the love of God, it is ridiculous to call QTGMC a deinterlacer, or more accurately a deinterlacing filter.

QTGMC is a script that calls a bunch of filters, including deinterlacers, noise removers, and sharpeners.

There is nothing QTGMC does that you can't do with an Avisynth script that calls and configures the the specific filters you want manually.

All QTGMC did was make it easy for people not experienced in writing scripts to work with video.

If i posted a poll that asked what is the best encoder, x264, x265. vp9, svt-av1 or Handbrake, what would you vote?

Would you say Handbrake or would you say it's a nonsensical poll?

[Sharc]

Most of these questions about "what is best ...." usually lead to nowhere ......

[lollo]

For the love of God...

Fot the love of God, re-read what we wrote, and if you are able to do so, compare the code for Yadiff, BWDIF, Nnedi3(field=-2), MCBobU(), TDeint, TempGaussMC(), QTGMC

[_Al_]

QTGMC is a deinterlacer package. The main purpose besides some others is to deinterlace video so it looks as if it was never interlaced. Which is a problem because of field definition. If someone comes out and claims it is not a deinterlacer, because it is a bunch of other filters and algorithms, then he is ******, there might not be a polite equivalent for that word. He might call it "spaghetti monster destroyer" maybe, if that is more appropriate for him or whatever else. Lets rest of the world call it deinterlacer as a base what it does.

[Skiller]

Trivial semantics. What does it matter if you call QTGMC a deinterlacing filter or a script? What is essential to the user is what it does.
A script is a type of code, isn't it? Don't other deinterlacers use code? I don't think the way it is "packaged" should matter in this question.

[sophisticles]

Trivial semantics. What does it matter if you call QTGMC a deinterlacing filter or a script? What is essential to the user is what it does.
A script is a type of code, isn't it? Don't other deinterlacers use code? I don't think the way it is "packaged" should matter in this question.

It matters because QTGMC does no deinterlacing in and of itself, instead it calls the other options listed in the poll depending on the settings you choose.

As I said it's the difference between Handbrake and x264, x265. svt-av1 and vp9; Handbrake does no encoding in and of itself, it's a front end.

Which is what QTGMC is, a front end for actual deinterlacers, and that's why this poll is nonsense.

[lordsmurf]

Trivial semantics. What does it matter if you call QTGMC a deinterlacing filter or a script? What is essential to the user is what it does.
A script is a type of code, isn't it? Don't other deinterlacers use code? I don't think the way it is "packaged" should matter in this question.

It matters because QTGMC does no deinterlacing in and of itself, instead it calls the other options listed in the poll depending on the settings you choose.

What other deinterlacer packages exist? None, I believe.
So until we start to see more packages that creates a category, I see no issue calling QTGMC a deinterlacer.

... but noting I refer to it as a "deinterlace method".

[Skiller]

It matters because QTGMC does no deinterlacing in and of itself, instead it calls the other options listed in the poll depending on the settings you choose.

Yes, but it merely uses those for spatial interpolation. You make it sound like QTGMC has nothing to offer on it's own. Re-read what I wrote about how the core of QTGMC works. That's where the magic happens.

[Bwaak]

You make it sound like QTGMC has nothing to offer on it's own.

QTGMC exists within Avisynth universe, it is not shippable/packageable as a distinct product or library or driver. It is a helper script, and it is nothing without its parts. I mean, combining software from existing parts is how pragmatic software dev is done. One can compare RedHat to Ubuntu to Debian, but comparing RedHat to, say, postresql makes no sense.

[lollo]

For the next release, somebody should compile QTGMC.avsi into a QTGMC.dll plugin and distribute just that, giving an end to this silly story.

[Skiller]

+1


Anyways, back to topic (sort of), why would people deliberately want to do single-rate deinterlacing nowadays? Many years ago it was often the only choice due to hardware and/or software restrictions. But today? It's literally throwing away half the video! Don't do it unless there is very good reason for it.

[_Al_]

You make it sound like QTGMC has nothing to offer on it's own.

QTGMC exists within Avisynth universe, it is not shippable/packageable as a distinct product or library or driver.

It is used in vapoursynth too. And as soon as you can use it in vapoursynth you can just make binaries - windows executable file - from python scripts and dependencies (directory with dll's, other scripts, other packages) . You just run a software where a file could be an argument or just using a gui, which is not really difficult to do in python. Or you can leave it as a module as well having it as a package. So it is "shippable and packabeable".

There is no problem at all to create an exe file, that can be run on any windows computer(guessing windows 7 and up) where you drop an argument on that exe (interlace video file) and it will use qtgmc and encode it. To visualize using a gui with fancy value options is just extra juice and effort while coding it, but same principle. You can use the same with avisynth and using windows batch, but there might be problems with windows system dll's, not sure. Using python you can just deviate into using gui with no problem.

[Bwaak]

This does not change the fact that QTGMC uses other "real" deinterlacers. This topic is about comparison of QTGMC with these deinterlacers, but you cannot compare a whole with its parts. I mentioned Linux distros above.

[_Al_]

I had QTGMC working on Ubuntu previously as well from ppa repository, wonderful djcj repository at that time. Not sure if anything exists up to date today.

How about bwdif, it is based on yadif and using w3fdif deinterlace method? So how it can be a deinterlacer when it uses other "real?" deinterlacer? It is really silly, it leads nowhere.

[Skiller]

but you cannot compare a whole with its parts.

I highly disagree with this. You still completely ignore the concept of the core, which is unique. I beginning to think you must be trolling!

[lollo]

Skiller, Bwaak is (almost) always trolling. Sometimes is funny, most of the times is not.

[Bwaak]

but you cannot compare a whole with its parts.

I highly disagree with this. You still completely ignore the concept of the core, which is unique. I beginning to think you must be trolling!

Not at all. QTGMC includes some of the deinterlacers it is being compared with, PLUS the core. In what situation it can be considered worse than an original deinterlacer? Maybe in case you want the least destructive deinterlacing so one can go back to interlaced, but as you said, the whole point of QTGMC was to take the result from a bobber, then to smooth it out, so there is no grounds for a comparison.

How about bwdif, it is based on yadif and using w3fdif deinterlace method? So how it can be a deinterlacer when it uses other "real?" deinterlacer? It is really silly, it leads nowhere.

I haven't look too deep into this, but is not bwdif a "same frame rate" deinterlacer, converting 25i/30i into 25p/30p? So it takes some functionality from yadif and throws away some other features and adds something else, it does not include yadif completely like in QTGMC case. A criteria to choose yadif vs bwdif could be whether one needs to double the frame rate; if not, then maybe bwdif can be better.

Skiller, Bwaak is (almost) always trolling. Sometimes is funny, most of the times is not.

Just because you cannot provide a definition of terms that you use does not mean I am always trolling

[Quint]

Interlaced content ---> BLACK BOX ---> Deinterlaced content.

If BLACK BOX uses other filters (that can deinterlace themselves) does not change the fact, that BLACK BOX is DE-INTERLACING, and thus BLACK BOX is a deinterlacer.

This whole discussion is senseless because of different goals, different sources, different opinions which result is better or worse, a. s. o...
But the part with QTGMC is the most senseless.

[Skiller]

but as you said, the whole point of QTGMC was to take the result from a bobber, then to smooth it out, so there is no grounds for a comparison.

That's not what I said and it's not how QTGMC works. You need to read properly.
I will write it again, very large this time.

QTGMC uses other bobbers for spatial interpolation only.

It does not take a bobber's output and "smooth" it.

I'm out of here.