Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion.

Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules.

 

Go Back   Doom9's Forum > Capturing and Editing Video > Avisynth Development

Reply
 
Thread Tools Search this Thread Display Modes
Old 2nd September 2021, 21:17   #1501  |  Link
ChaosKing
Registered User
 
Join Date: Dec 2005
Location: Germany
Posts: 1,795
So all fansubber?
__________________
AVSRepoGUI // VSRepoGUI - Package Manager for AviSynth // VapourSynth
VapourSynth Portable FATPACK || VapourSynth Database
ChaosKing is offline   Reply With Quote
Old 3rd September 2021, 12:05   #1502  |  Link
FranceBB
Broadcast Encoder
 
FranceBB's Avatar
 
Join Date: Nov 2013
Location: Royal Borough of Kensington & Chelsea, UK
Posts: 2,883
Quote:
Originally Posted by ChaosKing View Post
So all fansubber?
ROTFL
I've seen fansubber doing that in the 2006-2008 era.
Madness... and far too much free time those youngsters had back in the days...

(and I was "guilty" of being one of them back in the days :P )
FranceBB is offline   Reply With Quote
Old 9th September 2021, 15:41   #1503  |  Link
TheFluff
Excessively jovial fellow
 
Join Date: Jun 2004
Location: rude
Posts: 1,100
Even back when some anime nerds did do "manual" IVTC, they never actually went through every frame one by one. The typical workflow in tools like YATTA and its successors was to use pattern guidance; that is find sequences of frames with a consistent 3:2 cadence and then generate an override file for those sequences that forced that pattern. When you were lucky you only had like 3-4 pattern breaks for an entire episode (e.g. it was produced as progressive and telecined very late in the production process); if you were unlucky you had hundreds, but that's still far short of looking at every frame.
TheFluff is offline   Reply With Quote
Old 11th September 2021, 19:52   #1504  |  Link
Katie Boundary
Registered User
 
Katie Boundary's Avatar
 
Join Date: Jan 2015
Posts: 1,048
Quote:
Originally Posted by LigH View Post
Easy for humans. Quite a challenge for computer algorithms
LOL no. Interlacing is a mathematically very unique phenomenon that has nothing to do with image-recognition or any other human visual system bullshit. The key characteristic that makes it incredibly easy to distinguish from normal detail is that non-adjacent pixels are more similar to each other than adjacent pixels. In particular, for any given column of five pixels A, B, C, D, and E:

- pixels B and D will be more similar to each other than either one is to A, C, or E; and

- pixel C will be more similar to A and E than it is to B or D

This is literally all you need to distinguish interlacing from detail or from noise. If the problem seems "difficult", it's because certain people are ignoring differences between non-adjacent pixels and dicking around with multiplication and arbitrary user-defined threshold parameters, as if they were just copying and pasting from edge-detection algorithms. Hell, neither of TFM's combing metrics even bother to look at the differences between A and B, or between D and E, and Metric 1 doesn't bother looking at pixels A or E at all.

Quote:
Originally Posted by FranceBB View Post
Katie, you think they're different 'cause you "see" them as being different, but a computer doesn't see, a computer works with numbers.
No, they are different because they are MATHEMATICALLY different. In fact, I've sometimes had to use sharpening or other filters to reveal interlacing that was too subtle for the human eye to notice, but which a competently designed algorithm would have picked up easily.

Quote:
Originally Posted by FranceBB View Post
The whole frame is divided in blocks and macroblocks of a certain size, they can be 4x4, 8x8, 16x16 etc.
Macroblocks have nothing to do with deinterlacing because deinterlacing filters don't work on encoded video. They work on decoded video, which is a bunch of individual pixel values.
__________________
I ask unusual questions but always give proper thanks to those who give correct and useful answers.
Katie Boundary is offline   Reply With Quote
Old 12th September 2021, 12:04   #1505  |  Link
FranceBB
Broadcast Encoder
 
FranceBB's Avatar
 
Join Date: Nov 2013
Location: Royal Borough of Kensington & Chelsea, UK
Posts: 2,883
Quote:
Originally Posted by Katie Boundary View Post
Macroblocks have nothing to do with deinterlacing because deinterlacing filters don't work on encoded video. They work on decoded video, which is a bunch of individual pixel values.
By motion vectors I mean newly created Vectors to find differences on an uncompressed AV Stream frameserved by Avisynth. Vectors are created from scratch trying to compare blocks temporarily and thus finding out repeating patterns (this is how TDecimate gets rid of telecine for instance) but sometimes this doesn't work well and I've already explained why...
FranceBB is offline   Reply With Quote
Old 12th September 2021, 12:57   #1506  |  Link
StainlessS
HeartlessS Usurer
 
StainlessS's Avatar
 
Join Date: Dec 2009
Location: Over the rainbow
Posts: 10,980
If only we had some coders that were as brilliant as Katie, wouldn't that be something.
__________________
I sometimes post sober.
StainlessS@MediaFire ::: AND/OR ::: StainlessS@SendSpace

"Some infinities are bigger than other infinities", but how many of them are infinitely bigger ???
StainlessS is offline   Reply With Quote
Old 15th September 2021, 06:19   #1507  |  Link
Katie Boundary
Registered User
 
Katie Boundary's Avatar
 
Join Date: Jan 2015
Posts: 1,048
Quote:
Originally Posted by FranceBB View Post
By motion vectors I mean newly created Vectors to find differences on an uncompressed AV Stream frameserved by Avisynth. Vectors are created from scratch trying to compare blocks temporarily
Again, nothing to do with interlacing.

Quote:
Originally Posted by FranceBB View Post
repeating patterns (this is how TDecimate gets rid of telecine for instance) but sometimes this doesn't work well and I've already explained why...
YOU DON'T NEED TO LOOK FOR PATTERNS. YOU ONLY NEED TO COMPARE DIFFERENCES BETWEEN ADJACENT AND NON-ADJACENT PIXELS.

Quote:
Originally Posted by StainlessS View Post
If only we had some coders that were as brilliant as Katie, wouldn't that be something.
Coding = taking existing ideas and translating them into computer code.

I'm not good at that. I'm better at correcting stupid ideas before they reach the "translated into code" stage.
__________________
I ask unusual questions but always give proper thanks to those who give correct and useful answers.

Last edited by Katie Boundary; 15th September 2021 at 06:21.
Katie Boundary is offline   Reply With Quote
Old 15th September 2021, 10:55   #1508  |  Link
LigH
German doom9/Gleitz SuMo
 
LigH's Avatar
 
Join Date: Oct 2001
Location: Germany, rural Altmark
Posts: 6,753
So don't forget to consider all the possible kinds of motions that could appear in a video. There is not only "uniform linear horizontal", although it may be the most common. I am pretty sure one can find "academic examples" of motion which does not clearly calculate to lower differences for non-adjacent lines (like even pixel wide vertical).

I remember a more or less entertaining approach of (IIRC?) HDConvertToX: Let x264 encode a tiny crop one time with --tff and another time with --bff, and compare the resulting statistic outputs...
__________________

New German Gleitz board
MediaFire: x264 | x265 | VPx | AOM | Xvid
LigH is offline   Reply With Quote
Old 15th September 2021, 13:01   #1509  |  Link
FranceBB
Broadcast Encoder
 
FranceBB's Avatar
 
Join Date: Nov 2013
Location: Royal Borough of Kensington & Chelsea, UK
Posts: 2,883
Quote:
Originally Posted by LigH View Post
So don't forget to consider all the possible kinds of motions that could appear in a video. There is not only "uniform linear horizontal", although it may be the most common. I am pretty sure one can find "academic examples" of motion which does not clearly calculate to lower differences for non-adjacent lines (like even pixel wide vertical).
Yep, precisely.
That's exactly what I was trying to tell her...
It's not "that easy" nor "that simple" as she stated.
Anyway, I'm not gonna reply any further, I feel like I can't get my message through, but after all, I've always been terrible in talking with women even when it was weather-related chit-chat, let alone about encoding xD

Last edited by FranceBB; 15th September 2021 at 13:05.
FranceBB is offline   Reply With Quote
Old 15th September 2021, 16:55   #1510  |  Link
wonkey_monkey
Formerly davidh*****
 
wonkey_monkey's Avatar
 
Join Date: Jan 2004
Posts: 2,493
Quote:
Originally Posted by Katie Boundary View Post
In particular, for any given column of five pixels A, B, C, D, and E:

- pixels B and D will be more similar to each other than either one is to A, C, or E; and

- pixel C will be more similar to A and E than it is to B or D

This is literally all you need to distinguish interlacing from detail or from noise.
Nonsense. Interlaced video doesn't always meet these criteria and non-interlaced video doesn't always fail to meet them.

Besides, there's more to it than simply deciding (imperfectly) whether or not an individual pixel is "interlaced".

There's no perfect solution to this problem. Any attempt will involve compromises.

If you haven't even implemented your solution, how do you know it's better than any other?
__________________
My AviSynth filters / I'm the Doctor
wonkey_monkey is offline   Reply With Quote
Old 15th September 2021, 19:17   #1511  |  Link
Katie Boundary
Registered User
 
Katie Boundary's Avatar
 
Join Date: Jan 2015
Posts: 1,048
Quote:
Originally Posted by LigH View Post
So don't forget to consider all the possible kinds of motions that could appear in a video.
Motion has nothing to do with interlacing.
__________________
I ask unusual questions but always give proper thanks to those who give correct and useful answers.
Katie Boundary is offline   Reply With Quote
Old 15th September 2021, 19:35   #1512  |  Link
videoh
Useful n00b
 
Join Date: Jul 2014
Posts: 1,667
Quote:
Originally Posted by Katie Boundary View Post
Motion has nothing to do with interlacing.
Of course it does. If there is no movement between the fields then there will be no combing and no need for deinterlacing.
videoh is offline   Reply With Quote
Old 16th September 2021, 01:22   #1513  |  Link
Katie Boundary
Registered User
 
Katie Boundary's Avatar
 
Join Date: Jan 2015
Posts: 1,048
Quote:
Originally Posted by videoh View Post
Of course it does. If there is no movement between the fields then there will be no combing and no need for deinterlacing.
Unless it's around a scene change. Or it's a fade. Or there's literally any other change in the image that isn't caused by motion (someone turns a light on, for example).
__________________
I ask unusual questions but always give proper thanks to those who give correct and useful answers.
Katie Boundary is offline   Reply With Quote
Old 16th September 2021, 02:44   #1514  |  Link
videoh
Useful n00b
 
Join Date: Jul 2014
Posts: 1,667
Sure, those qualify as "motion between fields" too. But you made the false claim that motion has nothing to do with interlacing. What matters is whether the picture content changes across a field pair. Real motion is certainly a major instance thereof, and that explains why motion-adaptive methods can work better.

Last edited by videoh; 16th September 2021 at 02:48.
videoh is offline   Reply With Quote
Old 16th September 2021, 09:57   #1515  |  Link
LigH
German doom9/Gleitz SuMo
 
LigH's Avatar
 
Join Date: Oct 2001
Location: Germany, rural Altmark
Posts: 6,753
"Interlacing" means that two fields of a frame have been shot at different moments of time. If there is no motion, there is no difference to a progressive frame. "Combing" (while watching a whole frame) gets only visible for areas in motion because their content has different locations at different moments of time.
__________________

New German Gleitz board
MediaFire: x264 | x265 | VPx | AOM | Xvid
LigH is offline   Reply With Quote
Old 16th September 2021, 10:08   #1516  |  Link
StainlessS
HeartlessS Usurer
 
StainlessS's Avatar
 
Join Date: Dec 2009
Location: Over the rainbow
Posts: 10,980
Quote:
Originally Posted by FranceBB View Post
I've always been terrible in talking with women even when it was weather-related chit-chat, let alone about encoding xD
Me too, always seems like they would fail in a Turing Test :- https://en.wikipedia.org/wiki/Turing_test
EDIT: Or do I mean pass, or even Indeterminate.
__________________
I sometimes post sober.
StainlessS@MediaFire ::: AND/OR ::: StainlessS@SendSpace

"Some infinities are bigger than other infinities", but how many of them are infinitely bigger ???

Last edited by StainlessS; 16th September 2021 at 10:14.
StainlessS is offline   Reply With Quote
Old 18th September 2021, 03:36   #1517  |  Link
Katie Boundary
Registered User
 
Katie Boundary's Avatar
 
Join Date: Jan 2015
Posts: 1,048
Quote:
Originally Posted by videoh View Post
Sure, those qualify as "motion between fields" too.
Linguists and physicists, and pretty much the entire english-speaking world, disagree with you.

Quote:
Originally Posted by videoh View Post
But you made the false claim
There was nothing false about it. You can have motion without interlacing (it's called "progressive" video) and I already gave several examples of visible interlacing without motion.

Quote:
Originally Posted by videoh View Post
What matters is whether the picture content changes across a field pair.
Agreed. And that's not the same thing as motion.

Quote:
Originally Posted by LigH View Post
"Combing" (while watching a whole frame) gets only visible for areas in motion
Or if the picture changes for any other reason.
__________________
I ask unusual questions but always give proper thanks to those who give correct and useful answers.
Katie Boundary is offline   Reply With Quote
Old 18th September 2021, 11:47   #1518  |  Link
Frank62
Registered User
 
Join Date: Mar 2017
Location: Germany
Posts: 234
Interlacing or even combing has (directly) nothing to do with motion.
The detection of interlacing and combing has.
Frank62 is offline   Reply With Quote
Old 18th September 2021, 12:14   #1519  |  Link
wonkey_monkey
Formerly davidh*****
 
wonkey_monkey's Avatar
 
Join Date: Jan 2004
Posts: 2,493
Quote:
Originally Posted by Katie Boundary View Post
Linguists and physicists, and pretty much the entire english-speaking world, disagree with you.
This isn't a linguistic or physics forum. We have our own vernacular that doesn't always align with general usage, as does any field, and it's clear in this case that everyone except you is quite comfortable using the word "motion" to describe changes in pixel values. If you insist on being so pedantic (if that's even the right word when someone is incorrect), motion is at the very least one of the main causes of combing in interlaced video, so the claim that "motion has nothing to do with interlacing" is patently ridiculous in any case.

You might as well claim that there is no such thing as motion in a video at all, because it's just a collection of pixels changing over time.
__________________
My AviSynth filters / I'm the Doctor

Last edited by wonkey_monkey; 18th September 2021 at 15:39.
wonkey_monkey is offline   Reply With Quote
Old 19th September 2021, 05:28   #1520  |  Link
Katie Boundary
Registered User
 
Katie Boundary's Avatar
 
Join Date: Jan 2015
Posts: 1,048
Quote:
Originally Posted by Frank62 View Post
The detection of interlacing and combing has.
Nope. Detection only requires that the two fields be from slightly different images.
__________________
I ask unusual questions but always give proper thanks to those who give correct and useful answers.
Katie Boundary is offline   Reply With Quote
Reply

Tags
tdeint, tivtc

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +1. The time now is 16:16.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2024, vBulletin Solutions Inc.