Log in

View Full Version : Deinterlacing advice, which looks better?


VictiniSter
3rd March 2023, 04:20
Not sure if this is the correct place to post this but here goes.


So the reason I'm making this post is cause I am visually impaired so it's hard for me to tell what to look out for when comparing deinterlacing filters and their results as well as details like artifacts in general.

I have a video that is recorded in 1080i50 I would like to deinterlace into 720p50. Based off of some research I've largely used QTGMC for deinterlacing to 720p50 and NNEDI3 for single frame rate deinterlacing.

However it was suggested to me that I try using yadifmod2 as QTGMC can introduce shimmering into double frame rate video (or at least I think this is called shimmering, I'm referring to when you have something static like a logo that appears to be quickly moving up and down each frame)

Here's a 10 minute sample that was deinterlaced using QTGMC.
https://mega.nz/file/wsQDyA6B#StEbmp-0r0TZotT6lZvhH_PAjDspggHYySPE0XTCUzA


Here's the QTGMC call in the script, note I'm specifying the values of Rep0 and Rep2 in an effort to reduce shimmering without losing quality.

QTGMC(preset="Slower", Rep0=0, Rep2=0, EdiThreads=4)


Here's a 10 minute sample deinterlaced with yadifmod2

https://mega.nz/file/V1JCiKqJ#rKcsQTha3_umpKR-iBR_kAMJzJmzumioLnlPw5OwpC4

Here's the yadifmod2 call.


yadifmod2(mode=1)


And here's a 10 minute sample of the original source (ignore the FPS reported, that's just cause I trimmed the sample from the original source using LosslessCut)

https://mega.nz/file/ZtpEAZTI#gF6MB2JVOL_rGAVHEigdYRdG_uTtche9lfzDefBzt-o

So which of these looks better in your opinion (as in more visually appealing, not necessarily more accurate to the source)?

Katie Boundary
20th April 2023, 16:08
10-minute clips of the original, unfiltered, unedited video are good for "my source is pathological, wat do?" sorts of questions.

For side-by-side "what looks better?" questions, you're better off posting still images, because we're just gonna go through the videos frame-by-frame and pause them anyway.

Sharc
20th April 2023, 20:31
@VictiniSter:
A comparison should be based on equal bitrates (=equal file sizes), otherwise you won't compare apples with apples.
In your case the video bitrates of the 2 encodes differ by 30%. A comparison is pointless.

Selur
21st April 2023, 17:59
Am I missing something? Looking at the file without deinterlacing, I see no combing artifacts,... so why deinterlace something that isn't interlaced?
(qtgmc will probably lose quality wise when applied to a progressive source,...)

Sharc
21st April 2023, 19:42
The original is a mix of progressive and interlaced (possibly telecined - I didn't really check) video.

Selur
22nd April 2023, 06:30
Sharc is right. => Try TFM() instead of deinterlacing.

VictiniSter
6th January 2024, 19:12
I know this is 9 months later but thanks, I now see what you mean when you say it's a mix of progressive and interlaced. Now I gotta figure out how to only deinterlace the frames that need it (mainly cause i wanna keep the doubled frame rate otherwise I would just stick with TFM for the whole thing).

Sharc
7th January 2024, 09:47
It's a mix of interlaced, progressive and some animations with blended fields. I would probably just use QTGMC() when you want to preserve the motion of the interlaced stuff.

j7n
8th January 2024, 04:27
If I had a CPU to spare, I'd use QTGMC with SourceMatch=3 and Lossless=2 on interlaced or mixed content with potentially blended frames. If the video turns out to be mostly not interlaced, it can be decimated with srestore() with little loss of sharpness. I've heard people say that the output of QTGMC is too smooth. But some smoothing inevitably tends to happen at the encoding stage.