Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Recommended settings for optimal image quality #77

Open
ChristianWaechter opened this issue Jul 31, 2023 · 2 comments
Open

Recommended settings for optimal image quality #77

ChristianWaechter opened this issue Jul 31, 2023 · 2 comments

Comments

@ChristianWaechter
Copy link

Hello,

I have a question when it comes to the different presets for QTGMC+. In the past, some of the presets of QTGMC smoothed the image a bit too much, as seen here: http://macilatthefront.blogspot.com/2021/03/lets-test-qtgmc-settings.html

This comparison has been done with 3.376 of QTGMC without +.

How is the philosophy of QTGMC+ with the image quality? Can one safely use the Placebo preset and get the overall best quality or is it required to test various settings? E.g. I would expect that the Placebo preset also uses SourceMatch=3 and so on.

How about the additional EdiModes considering quality. The Placebo defaults to BWDIF+BOB+NNEDI3 as far as I can see, so I would assume this delivers the best quality. Or are there also any downsides like too much smoothening and so on?

Is there even a guideline that lists the different characteristics/(dis)advantages of theses additional QTGMC+ EdiModes?

Thank you very much :)

@Dogway
Copy link
Owner

Dogway commented Jul 31, 2023

QTGMC+ is very source dependent, so for example, is the source clean or grainy?

For clean sources like current 1080i captures a recent call I used was:

QTGMCp(thSAD1=640, thSAD2=256, Preset="Slower", SourceMatch=0, MatchPreset="Slower", MatchPreset2="Slower", MatchEnhance=0.5, Lossless=0, EZDenoise=0.0, EZKeepGrain=1.0, NoisePreset="Faster", gpuid=0, threads=4)

Placebo is almost never worth it. If aiming for detail preservation I'd use SourceMatch no higher than 1, and keep lossless at 0, over this it leaves combing.

On EdiModes, yes, it's easy to get fooled due to the writting, but mind line 245, all presets use "NNEDI3" as EdiMode. The ones listed in line 244 is when moiree is enabled. This can give you a less blurry output but it's prone to artifacts.

All in all quality wise QTGMC+ should give you a little more quality than v3.376 for a few reasons. Luma_Rebuild prefilter, base bob with BWDIF() instead of Bob() and merged operations (less quantization steps to 8-bit). You can also run it in HBD for extra quality. The rest of the EdiModes are more like a quality/performance compromises, in the slower ones is more like a artifact/quality ratio. You can run "BWDIF+BOB+NNEDI3" but does it remove all combing? That's why by default all presets default to "NNEDI3".

There are more improvements on the MVTools side, like rfilter, increased chroma sampling, and bugfixes as you can read in the changelog.

@ChristianWaechter
Copy link
Author

ChristianWaechter commented Aug 1, 2023

I would not say that my source is grainy, but it is SD. I'm currently working with old Video8 tapes that are already 35 years old and some newer miniDV tapes (~20 years old). Resolution is 720 x 576 (PAL). Here is an example frame that has been deinterlaced just with QTGMCp(Preset="Slow")

Full Processing Script 2 64bit 01

You can run "BWDIF+BOB+NNEDI3" but does it remove all combing?

That is an interessting question. Let me ask, does "BWDIF+BOB+NNEDI3" remove more combing than "NNEDI3" alone? I know that there will alway be some artifacts, so I'm more looking for the method that shows the fewest.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants