Why deinterlacing is (tried) done, as it is impossible?

luquinhas0021 · 26th December 2015, 18:07

Well, you know there's two types of deinterlacing: those is done in record and those is done in post-production. Those is maden in record has one frame scaned twice, two times.
In post-production interlaced video, the software generates two fields, one with odd lines, another with even lines. Deinterlace it is tottaly possible: just join two fields (Weaving deinterlacing).
But, in record interlaced content, the things starts to be hard. You should know that, in this type of interlacing, in first time, only the odd lines writen in sensor are scaned; on second time, only even lines are scaned (But enter light in whole sensor). How this scans are maden in different times, is problably that will have move displacement, i.e, two frames inside only one.
I don't believe in possibility of this type of deinterlacing, although there are so much tecnhiques in order to try do this. As there are two movements, any deinterlacing technique or causes artifacts on "progressive" image, or, in case of motion compensation deinterlacing, makes loss some movement.
So you ask me: eaht are your solution? First, try to make weaving deinterlacing. If combing artifacts appears, I suggest separate the fields of video, upscale one by one with your liked algorithm and apply motion compensation two by two fields (Or new frames) for recovering the details lost in interlacing process. In Avisynth:

(Avi or DirectShow)Source("...")
SeparateFields
(Some upscaling algorithm)(x, 2y)
(Some motion compensation algorithm, like MVTools)

feisty2 · 26th December 2015, 18:36

I hate to be mean, but
gonna summarize this whole long statement with 2 words:
"Common Sense"

luquinhas0021 · 26th December 2015, 18:38

It isn't. I have a Brazilian friend who is specialist in this. He sad it I sad isn't foolish

ndjamena · 26th December 2015, 18:58

TFM with QTGMC used as a deinterlacer source?

I gave up on that.

The first episode of Buffy had a panning fade-in that no combing detector would discover without making it sensitive enough that I might as well have skipped the TFM.

Or are you suggesting ignoring the interlaced flag first, then try QTGMC if that doesn't work?

wonkey_monkey · 26th December 2015, 22:00

Quote:

So you ask me: eaht are your solution? First, try to make weaving deinterlacing. If combing artifacts appears, I suggest separate the fields of video, upscale one by one with your liked algorithm and apply motion compensation two by two fields (Or new frames) for recovering the details lost in interlacing process.

Isn't that exactly what QTGMC (and mcbob before it) does?

If a portion of the picture is static, keep it as it is. If it isn't, do some motion detection and see if the "moved" details can be moved back to replace the missing information. Otherwise, fall back to interpolation.

No offence luquinhas0021, but for almost every problem you boldly atttempt to tackle, better minds than yours have already spent far more time on the problem than you ever will.

luquinhas0021 · 27th December 2015, 03:51

david, honestly, i didn`t know how qtgmc works, and, even more honestly, i ain`t obbligated to know how this or that algorithm performs deinterlacing. I, no lie, know the deinterlacing techniques, however I still didn`t link that script to that technique. Ben, unless qtgmc turn x fields per second in x frames per second, qtgmc isn`t equal to my suggestion. The most of deinterlacing transforms x fields per second in (x/2) frames per second. I didn`t propose this type of deinterlacing.

ndjamena · 27th December 2015, 04:15

Quote:

Originally Posted by luquinhas0021

david, honestly, i didn`t know how qtgmc works, and, even more honestly, i ain`t obbligated to know how this or that algorithm performs deinterlacing. I, no lie, know the deinterlacing techniques, however I still didn`t link that script to that technique. Ben, unless qtgmc turn x fields per second in x frames per second, qtgmc isn`t equal to my suggestion. The most of deinterlacing transforms x fields per second in (x/2) frames per second. I didn`t propose this type of deinterlacing.

-edit- god, he hadn't even tried QTGMC before posting.

vivan · 27th December 2015, 11:38

Quote:

Originally Posted by luquinhas0021

david, honestly, i didn`t know how qtgmc works, and, even more honestly, i ain`t obbligated to know how this or that algorithm performs deinterlacing.

No, you have to do at least some research before creating topics.

Quote:

Originally Posted by luquinhas0021

The most of deinterlacing transforms x fields per second in (x/2) frames per second. I didn`t propose this type of deinterlacing.

Why are you claming this even though you've done zero research?

wonkey_monkey · 27th December 2015, 14:10

Quote:

Ben, unless qtgmc turn x fields per second in x frames per second, qtgmc isn`t equal to my suggestion.

It can do exactly that.

Quote:

The most of deinterlacing transforms x fields per second in (x/2) frames per second.

Some deinterlacers do this by default, but they almost all have the ability to reconstruct x frames per second.

And if, for some reason, they don't, then you can fake it simply by calling it twice on field-swapped clips.

ndjamena · 27th December 2015, 14:27

Can you at least test these things before commenting about them. You're confusing me by claiming QTGMC doesn't do what it's designed to do.

What you're talking about is called a bobber. The most basic research would have told you that.

Here is an (incomplete) list of avisynth deinterlacing filters:

http://avisynth.nl/index.php/Externa...#Deinterlacing

Anything that mentions the word bob (and most of the others as well) do exactly what you're saying, in fact I'm not sure where you got the idea that most deinterlacers output at single rate. They don't.

26th December 2015, 18:07	#1 \| Link
luquinhas0021 The image enthusyast Join Date: Mar 2015 Location: Brazil Posts: 270	Why deinterlacing is (tried) done, as it is impossible? Well, you know there's two types of deinterlacing: those is done in record and those is done in post-production. Those is maden in record has one frame scaned twice, two times. In post-production interlaced video, the software generates two fields, one with odd lines, another with even lines. Deinterlace it is tottaly possible: just join two fields (Weaving deinterlacing). But, in record interlaced content, the things starts to be hard. You should know that, in this type of interlacing, in first time, only the odd lines writen in sensor are scaned; on second time, only even lines are scaned (But enter light in whole sensor). How this scans are maden in different times, is problably that will have move displacement, i.e, two frames inside only one. I don't believe in possibility of this type of deinterlacing, although there are so much tecnhiques in order to try do this. As there are two movements, any deinterlacing technique or causes artifacts on "progressive" image, or, in case of motion compensation deinterlacing, makes loss some movement. So you ask me: eaht are your solution? First, try to make weaving deinterlacing. If combing artifacts appears, I suggest separate the fields of video, upscale one by one with your liked algorithm and apply motion compensation two by two fields (Or new frames) for recovering the details lost in interlacing process. In Avisynth: (Avi or DirectShow)Source("...") SeparateFields (Some upscaling algorithm)(x, 2y) (Some motion compensation algorithm, like MVTools) __________________ Searching for great solutions

26th December 2015, 18:38	#3 \| Link
luquinhas0021 The image enthusyast Join Date: Mar 2015 Location: Brazil Posts: 270	It isn't. I have a Brazilian friend who is specialist in this. He sad it I sad isn't foolish __________________ Searching for great solutions Last edited by luquinhas0021; 26th December 2015 at 18:41.

27th December 2015, 03:51	#6 \| Link
luquinhas0021 The image enthusyast Join Date: Mar 2015 Location: Brazil Posts: 270	david, honestly, i didn`t know how qtgmc works, and, even more honestly, i ain`t obbligated to know how this or that algorithm performs deinterlacing. I, no lie, know the deinterlacing techniques, however I still didn`t link that script to that technique. Ben, unless qtgmc turn x fields per second in x frames per second, qtgmc isn`t equal to my suggestion. The most of deinterlacing transforms x fields per second in (x/2) frames per second. I didn`t propose this type of deinterlacing. __________________ Searching for great solutions

26th December 2015, 18:36	#2 \| Link
feisty2 I'm Siri Join Date: Oct 2012 Location: void Posts: 2,633	I hate to be mean, but gonna summarize this whole long statement with 2 words: "Common Sense"

26th December 2015, 18:58	#4 \| Link
ndjamena Registered User Join Date: Sep 2012 Posts: 366	TFM with QTGMC used as a deinterlacer source? I gave up on that. The first episode of Buffy had a panning fade-in that no combing detector would discover without making it sensitive enough that I might as well have skipped the TFM. Or are you suggesting ignoring the interlaced flag first, then try QTGMC if that doesn't work?

27th December 2015, 14:27	#10 \| Link
ndjamena Registered User Join Date: Sep 2012 Posts: 366	Can you at least test these things before commenting about them. You're confusing me by claiming QTGMC doesn't do what it's designed to do. What you're talking about is called a bobber. The most basic research would have told you that. Here is an (incomplete) list of avisynth deinterlacing filters: http://avisynth.nl/index.php/Externa...#Deinterlacing Anything that mentions the word bob (and most of the others as well) do exactly what you're saying, in fact I'm not sure where you got the idea that most deinterlacers output at single rate. They don't.