News:

--

Main Menu

XviD aspect bugs ?

Started by kapetr, January 19, 2015, 12:00:17 PM

Previous topic - Next topic

AQUAR

#30
The merit of this topic relates to the fact that there is only one PAR that correctly specifies the DAR of anamorphic PAL.
By DAR we are talking about projecting the visual content in a manner that shows objects in the image in their correct display aspect.

This topic starts with the title "ADM has an Xvid aspect ratio bug".
There is no bug, ADM correctly adds meta data for projecting the correct DAR of anamorphic PAL.

People want to believe that the DAR of a PAL source must be a perfect match for a screen aspect (canvas!) of 16:9.
Logical conclusion - Yes.  Correct - Yes AND No.
Lots of information do quote a PAR of 64/45 for DV-PAL (google will confirm!).
Why? because it redefines the projection of the SAR so it fits the 16:9 screen (by slighly distorting the DAR!).

Impact of fiddle - not perceptible for movies (other fiddles happen too eg, speeding up the sound for film on PAL).
But if you were a draftsman - you would get fired for using that PAR.

The only person that made a reaction on this topic and that has understood the issue of this topic is Jan.

I don't like to argue, and when I said so @kapetr told me to stop arguing and go elsewhere.
Am I peeved because of that?  Yes.
Am I wrong in my perception of this issue? Leave that up to others to judge (in fact - who cares for transcoding movies!).
Is this an "aspect Xvid BUG" in ADM? Everything that disagrees with enduser expectations here seems to be presented in the form of "it is a BUG".

My approach to dealing with PAL (16:9):
Happy to apply either of the common PAR values (64/45 or 16/11) for movies I just watch and forget.
For important or profesional stuff, I keep the DAR, meaning either:
1) It fits (usually analogue PAl) (when Xvid was invented!) or
2) It needs vertical letterbox scaling or cropping (and maybe a bit of both for container compliance).

Hopefully some readers will appreciate my perception of this item.
In any case, take it or leave it, as I don't care anymore about elaborating or contributing to this topic because of the negative attitudes. 

poutnik

#31
I can agree with that.
With the point ADM is not professional tool and  its usage does not always follow professional ways.
Small distorsion of intended DAR are fully acceptable.

OTOH, as extreme, I cannot withstand way of my parents always watching 4:3 stuff on 16:9 screen with nice Sun ellipses and fat people.

As curiosity, I sometime apply realtime Avisynth postprocessing using warpedresize function of simpleresize plugin,
http://forum.doom9.org/showthread.php?p=1619712#post1619712
partically correcting aspect ratio issues at such viewing 4:3 on 16:9.

It works by nonlinear transformation of X and Y coordinates.
Effects on both axis are adjustable and independent, key is not to overkill.

Instead of massive effect of full correction for central region in single dimension, leading to unpleasant distortion,
partial and the opposite corrections in both dimensions are used and effect is fine.

Horizontally, It little shrinks central screen region, while little expands border screen regions.
At the same time...
vertically, It little expands central screen region, while little shrinks border screen regions.

So if circle of diameter 30 is now 40x30 ellipse on wide screen centre. with aspect 4:3
it will do from it e.g. ellipse 36x33, with aspect 12:11


AQUAR

Some widescreen TV's and a few media players had a similar, non linear strech from the centre point, function.
I've seen that avisynth function mentioned before (not used it myself!).
AFAIR it sectionalized areas of each frame and then applied Lanczos resize at various strenghts to them.

Easy introduction to using avisynth (windows PC!) with ADM is by way of Mulders AVS-Proxy_GUI.
Coupling Avisynth to Avidemux opens up some very interesting posibilities for video manipulation.

kapetr

Quote from: AQUAR on January 22, 2015, 03:05:10 AM
This topic starts with the title "ADM has an Xvid aspect ratio bug".
There is no bug, ADM correctly adds meta data for projecting the correct DAR of anamorphic PAL.
..
Lots of information do quote a PAR of 64/45 for DV-PAL (google will confirm!).
Why? because it redefines the projection of the SAR so it fits the 16:9 screen (by slighly distorting the DAR!).

The target is not to fit video into 16:9 TV screen with "slighly distorting the DAR". Not!
The target is to display video in correct DAR of source - circle must looks as circle. That is correct way to display video.

=> ADM DO NOT "correctly adds meta data for projecting the correct DAR of anamorphic PAL" - while it ignores SAR and uses only one PAR for all possible SARs of 16:9 original video.

As I wrote - for DVB ("digital PAL") the SAR can be for 16:9 DAR of original source (!) 720x576, 704x576, 544x576, 480x576.

If you thing, that is it correct to display original 16:9 DAR source transmitted in 480x576 resolution using (as you thing) "the only one correct 16:11 PAR for anamorphic PAL" - which results in video with aspect ratio 698:576 (displayed DAR), then you have really - really very low demands on watching videos.

Circle with  diameter of 30 will be displayed as ellipse 20x30. [30*698/1024] No thanks!
It is for you "slightly distorted" ... Not for me. Not for me.

poutnik

Quote from: AQUAR on January 22, 2015, 08:48:25 AM
Some widescreen TV's and a few media players had a similar, non linear strech from the centre point, function.
I've seen that avisynth function mentioned before (not used it myself!).
AFAIR it sectionalized areas of each frame and then applied Lanczos resize at various strenghts to them.

That Panorama function the Doom9  the thread was about tried to achieve the effect by quantized stepwise way.
While AVS Script Function  PanoramaWR/WR2, based on WarpedResize function of Simpleresize plugin does it by continous way.

Yes, Avisynth provide to ADM the whole ocean of options and possibilities.
Not sure if Avxsynth or Vapoursynth are mature enough to be used for ADM in Linux.


AQUAR

@ poutnik,

Its an interesting function for sure - going to try it on a sample just to see the effect.

@ kapetr,
Obviously you are the aspect expert.
Therefore, I wish you the best of luck in convincing the ADM author to fix this BUG.
Nothing more to say!


kapetr

#36
Exactly what I did expect.
You have theory - that really is OK - you are using anchored http://en.wikipedia.org/wiki/Standard-definition_television#Pixel_aspect_ratio
from <Jan>.

Quote
Note that the actual image (be it 4:3 or 16:9) is always contained in the center 704 horizontal pixels of the digital frame, regardless of how many horizontal pixels (704 or 720) are used. In case of digital video line having 720 horizontal pixels, only the center 704 pixels contain actual 4:3 or 16:9 image, and the 8 pixel wide stripes from either side are called nominal analogue blanking for horizontal blanking and should be discarded before displaying the image.

It there would be only 720x576 and 704x576  possibilities then these theory could be maybe true. And you could be right!
720x576 16:9 video could be "correctly with 16:11 PAR" cropped to central 704 points and then displayed as 1024x576 DAR without be distorted.
Or it could be 1048x576 - but displayed also as 1024x576 (resized 720->1024) with small distortion to fit in 16:9 TV/monitor.

But the problem I have with you is, that I had many times present arguments/evidences to refute this theory (IMO):
- the other SARs.

I would expect from you 2 possible reactions:

1. explanation of how to get 704 central points from horizontal 544 or 480 points (or even 352 ! on DVD). In other worlds, how with these theory (using  "the only correct 16:11 PAR") display 544x576 and 480x576 (and 352x576) with acceptable/no  distortion as 16:9 original DAR video.
[to show you are right and I'm wrong]

or - if you can not, then you should be man enough to say:

2. I was wrong. This theory fails.

But unfortunately you are ignoring arguments ("Ostrich strategy" - see no evil, hear no evil), which could show you are wrong.
You have never responded to these arguments.
Just sarcasm to say to me: "You're an idiot, to which does not make sense to explain something."


>:(

zakk

Good luck kapetr (see private message).

poutnik

Aside of fact, what side is closer to truth,

rather cooperation than confrontation spirit suits to all involved parties in all discussions,
with focus to properly analyse point of view of the other side.

As they may just see the opposite side of the truth, that we at the moment cannot or do not want to see.
Realization of that may teach something all of them.

As if we had a ball, one hemisphere black, one hemisphere white,
one can say the ball is black, while the other NO,NO, the ball is white.

And both are true.


AQUAR

#39
@ zakk - private message but no public contribution (I can only guess based on history!)

@ poutnik - more like a case of which shade of grey is darker when they are subjectively the same!
Only recently did we have a nice discussion that did not degenerate because of divergent opinions.
It's all in the approach of the OP in steering their thread and with respect for other opinions.

@ kapetr - I was just focussed on PAL (16:9) in my reaction - the rest of your SAR's are fluff to obfuscate.
I told you my approach - again it doesn't meet your expectation or understanding - and so the crap goes on.

Staying with PAL (16:9) in the list of Xvid configuration options.
(Configuration options whose roots are defined by the Phase Alternating Line (PAL) encoding system for analogue television)

I perfectly understand Your perception viz:
You want a PAR to fit all of SAR into (16:9) frame and KEEP DAR correct.
That flows on from the idea that a (16:9) source must be a perfect fit for (16:9) screen and therefore has to be coupled to a PAR to suit.
Already said that is a logical conclusion but its not always so (highlighted by the way these older codecs are optioned!).

Case 1:
For DV-PAL using PAR 64:45 it will fit 16:9 frame.
Point off difference is.
You maintain that the DAR will be correct.
I say its slightly altered, the DAR will be slighly off (about 2.5%).
For DVD-PAL with SAR 720/576 the actual defined "square equivalent" is 1050/576.
For important stuff, its a case of deinterlace, resample and crop (already mentioned).. 
Mini explanation - - -
The PAL in the DV-PAL signifies the compatibiliy for use with PAL TV's.
The virtual PAR of PAL TV's doesn't change because of resolution changes in the SAR (again just focus on the 16:9 case!).
For digital TV's you need appropriate processing from THAT baseline (another can of worms topic!). 

Case 2:
For older "analogue" SAR using PAR 16:11 it will fit 16:9 frame.
You maintain that the DAR will be correct.
I never disagreed (just got the cases back to front in one of my reactions).
Mini explanation - - -
The virtual pixel resolution was calculated based on the sweep rate and modulation capability of PAL tv's.
And only in terms of the useable visible content in such signals (no overscan, teletext etc).
Result 702 : 576 resolution described by virtual pixels with a PAR of 16:11 (approximately).
It doesn't change because DV-PAL has a SAR with more resolution (720 : 576) (ie compatibility is maintained!).

Now, I never called you an idiot and never said your approach is wrong.
I just didn't want to argue about it anymore to avoid what this thread has now turned into.
In fact, I said right at the start that I hope someone else will give a proper explanation.
Why - because this kind of argie bargie tends to develop with complicated topics that start with "Its a BUG".
And usually egged on by certain individuals.

AGAIN, we are looking at this issue from a different baseline and hence our perception of how to use this Xvid configuration option is also different.
For the record, I don't need a wiki theory link provided by Jan to anchor my perception, it was stated here before that.
Perhaps You might do well to digest that material as it is obviously also at odds with your claim.
 
Your argument is based on what is widely flaunted as the PAR to use for DV-PAL.
That information isn't the whole truth, so look at that if you are so pedantic about DAR perfection.
Ironically - I am fine with "that PAR" as an extra option - already said I am happy to use either for movies.

I also said Xvid is an older codec that dates back to the time period when your perception actually fits the video stuff of that time.
ADM2.6 is geared up to work properly with the current crop of codecs, just has Xvid in cutdown configurability for transcoding from AVC.
I have suggested several times that you use ADM2.5 as it is better suited to your needs in terms of these PAR fiddles. 

IMHO,  I don't think the developer is going to bother with providing options on older codecs that represents such sudo updates (in X264 style) as you are asking for.  It is fine the way it is and is for the last time NOT A BUG. 

Finally (sigh!) DVB is digital broadcasting and nothing to do with PAL encoded broadcasting.
The mechanism for achieving 16:9 aspect from whatever SAR is achieved by scaling (again not interested in that tangent).

So! - one more insult from you (like the rude ostrich comparison) - and I will delete wholesale my reactions in this thread.
(if the administrator doesn't beat me to it by deleting the whole thread!)


poutnik

#40
Quote from: AQUAR on January 23, 2015, 02:34:03 AM

@ poutnik - more like a case of which shade of grey is darker when they are subjectively the same!
Only recently did we have a nice discussion that did not degenerate because of divergent opinions.
It's all in the approach of the OP in steering their thread and with respect for other opinions.

Who is without flaw, let throw a stone as the first...

It is perfectly normal to have different opinions, having different points of view and different preferences.
The key is the art to deal with it.  Sometimes is enough to try to stand myself at place of the others and look at it.

AQUAR

#41
@ poutnik
Tolerance is indeed a virtue.
Now, what do you do when the first stone has been thrown?

This item should really be a feature request, to add an extra option for the now popularised PAR for DV-PAL.
The Author/developers will decide that on case merit and priorities.

In the back of my mind I wonder why recoding with Xvid is used at all for DV-PAL, when X264 is more current for the all digital environment.
Only reason I can think of is playback without X264 decoding ability.

 


poutnik


Does it mean there was found anybidy without a flaw ? But if a stone is thrown, than learning to avoid it, or going away.
Flamewars do not serve any useful purpose.

If there is list of common predefied PAR, and option to define custom one, I do not not see any problem.

XviD: I guess there are many legacy HW players refusing H264.  I still see more XviD shared records than H264 ones.



kapetr

#43
Quote from: AQUAR
I was just focussed on PAL (16:9) in my reaction - the rest of your SAR's are fluff to obfuscate.
I told you my approach - again it doesn't meet your expectation or understanding - and so the crap goes on.

Everything what shows yours theory is wrong, is "fluff to obfuscate". Of course - what else to expect from you  :D
It is impossible to discuss  constructively with such people => this is my last reaction on yours posts.
What ever "smart" or rude you will say. END.

Quote
I perfectly understand Your perception viz:
You want a PAR to fit all of SAR into (16:9) frame and KEEP DAR correct.
That flows on from the idea that a (16:9) source must be a perfect fit for (16:9) screen and therefore has to be coupled to a PAR to suit.
Already said that is a logical conclusion but its not always so (highlighted by the way these older codecs are optioned!).

You do not understand - anything.

DAR is not about "fitting" somewhere.  It is not about what the display/monitor is: 5:4, 4:3, 16:10, 16:9, 2.35:1, ...
It is there - to tell to player (SW, HW, TV, ....) which aspect radio the source video has - to circle be circle on the attached monitor.
It must be: DAR=SAR*PAR - if not, then it is wrong (more ore less).

Only after that is cleared - it can be displayed on monitor - and it is another story - what will player do with the information about DAR of video. But this information must be correct. If not - the final result can not be correct anyway.

The player must e.g:
- taking into account PAR of monitor (e.g. 16:9 plasmas with 1024x768 resolution)
- taking into account users prefs (zooms, not-linear possibilities , ...)
- do resizing and cropping or adding black margins ...

All other what you say is again and again BLA-BLA about fitting something somewhere.
It makes no sense to comment again.


P.S: Thanks <zakk>, I understand.

zakk

Quote from: AQUAR on January 23, 2015, 02:34:03 AM
@ zakk - private message but no public contribution (I can only guess based on history!)
You alreay know what I think of your behavior in this forum.