why do many DVDs have 16px black border space @ left+right?

[lovelove]

Hi. I was wondering why many DVDs have a 16px black frame left+right (but not on top/bottom) ?

screenshot:
http://img847.imageshack.us/img847/4984/vlcsnap2011081111h42m37.jpg

Google doesn't know. The only reason I could think of, was that it would possibly get cut off on CRT screens ... but not on LED. And one would think current commercial DVDs would cater to the majority?

[konran]

Hi. The reason is based on different television broadcasting standards. In the very past, e.g. PAL had 704x576 format and changed that later to 720x576. When I remember correctly it is similar in NTSC broadcast area (704x480 vs. 720x480). In the early 2k's I had a Panasonic DVD/HDD recorder which always recorded at 704x576, so there are valid reasons for this format.

You may watch for "World's Television Systems" or "DVD Demystified" in your research.

[Coolpplse]

If I remember correctly its also because 16:9 480/576 material is not truly 16:9, theres a slight AR difference between actual 16:9 material/HD and Widescreen NTSC/PAL

[MrVideo]

Technically it is how the original NTSC, or PAL, video is converted to digital.

For NTSC, the 64 microsec active video line is supposed to be 704 pixels. But, when capturing for 720 pixel files, the scan width is widened a little bit to cover the front porch and back porch black. Look up the NTSC video spec to see what NTSC analog video looks like.

When converting the 720x480, ot 720x576, video for display on a digital display, the 8 pixels, on the left and right, need to be stripped and 704x480(576) video encoded for 4:3 (10:11 SAR) or 16:9 anamorphic widescreen (40:33 SAR) (12:11/16:11 for 704x576)

I'm assuming recoding to H.264 with x264.

With analog CRT screens, the overscan has nothing to do with those black bars, as described above. A DVD player, if designed correctly, will take the 720 pixels and place it in the active video line so that the center 704 pixels = the 64 usec.

So, yes, if manufacturers were to change to 704x480(576), then the digital displays would be more accurate as far as the aspect ratio is concerned. But that would require "fixing" all of the conversion gear that is used.

Plus, these days, movies and TV shows are all mastered in the HD realm (movies edited at 4k, 8k, or higher resolutions). They are then converted to MPEG-2 720x480(576) for release. No idea why the 8 pixel black is added when mastering when 704 would be better.

Oh, and the info can be found via Google. You just needed to use different search words. :D

[hello_hello]

So why do (some?) players (my Sony Bluray player connected via HDMI for example) always over-scan when playing non-DVD content on a digital display? Even when playing a 656x368 AVI it still over-scans. It doesn't seem to mess with the aspect ratio as it over-scans top and bottom as well as the sides, but it seems kind of redundant.

It does make me wonder about full HD content too. If the TV is 1920x1080 and so too is the video, and the player is over-scanning, what's the logic? Wouldn't it be better not to scale the video at all (the same way my PC doesn't)?

I only recently bought a Sony Bluray player, and while comparing the picture between it's 1080p MKV output and the PC's output (currently connected via the analogue PC input on the TV), the PC's picture quality looks better to me. Not by much, but I think it does look better (although I'm still fiddling to try to adjust the contrast and color etc to to match them (the TV doesn't have color adjustments when using the PC input, only brightness and contrast).

Edit: Well unnecessary over-scanning aside, the more videos I compared the more I realised the PC's analogue picture was always going to look more natural than the standalone player's HDMI picture, no matter what settings I fiddled with. Not a lot in it, but each time I ran two identical videos and switched between inputs, I could definitely see it.

[TheSkiller]

So why do (some?) players (my Sony Bluray player connected via HDMI for example) always over-scan when playing non-DVD content on a digital display?
I'm pretty sure it's not your Bluray player that is doing the overscan but your TV. New TVs do overscanning as well because if they didn't you would obviously have situations in which you see nasty borders and blinking white lines at the edges (depending on the content but with regular TV programs you would see this sooner or later).
Your TV does overscan on all of it's "video" inputs, that is, all the inputs exept the VGA input as that one is pretty much exclusively used for PCs where you don't want any overscan.
So yeah, if you play a 1920x1080 video via your Bluray player on a Full HD TV then you don't get 1:1 pixel mapping. Most TVs I've seen however allow the user to disable the overscanning with HD resolutions. Have a look.


@ MrVideo You're right but just for the record, the active picture is 52µs long for PAL (equals 702 pixels) and 52.6555µs for NTSC (equals 710.85 pixels), 64µs is the length of the whole scanline including blanking and color burst etc. ;) 702 is the exact number of pixels used to carry the active picture (for PAL) but for all practical applications even I use 704 as that is a nice resoluton to encode to.

[Ghitulescu]

It's strange that someone so knowingful as hello_hello asks such a basic question, which can be answered using wikipedia/google (for really, really basic terms) and a 3-steps approach.

@TheSkiller - it's always the TV the device in the chain that does the overscan, not the player. A player does only upscaling/downscaling, and only when it's needed.
Also a FullHD TV doesn't do overscan at 1920x1080 via HDMI, the earlier models did this however for mainly two reasons (1. to simplify the FW, 2. to allow the image to orbit, to prevent plasma burn-in effects). The LCD TVs are pixel-perfect from factory, to prevent an extremely visible loss in resolution (the scaling is more noticeable on a LCD than on a plasma). The overscan is always enabled (and cannot be disabled) on all analog inputs.

[hello_hello]

It's strange that someone so knowingful as hello_hello asks such a basic question, which can be answered using wikipedia/google (for really, really basic terms) and a 3-steps approach.

I thought I'd ask here instead and give you the opportunity to put your foot in your mouth.


Also a FullHD TV doesn't do overscan at 1920x1080 via HDMI, the earlier models did this however for mainly two reasons (1. to simplify the FW, 2. to allow the image to orbit, to prevent plasma burn-in effects).

Let me explain my last post slowly, in a 9-steps approach to make it easy for you.

1. I copy a 1080p video onto a USB stick.
2. I plug the USB stick into the Sony player.
3. I press play on the remote.
4. I pause the video.
5. I play the same video on the PC.
6. I pause it at the same place.
7. I then switch back to the HDMI input.
8. I compare the two pictures.
9. When I did the above, there was no doubt the video via the HDMI input was being over-scanned. I worked it out all by myself by looking at objects near the end of the screen and noting how close they were to it. Station logos make it really obvious.

The LCD TVs are pixel-perfect from factory, to prevent an extremely visible loss in resolution (the scaling is more noticeable on a LCD than on a plasma). The overscan is always enabled (and cannot be disabled) on all analog inputs.

Well despite the fact that you obviously think colouring in your words makes them true, the first thing I did after connecting the PC to the TV (51" Full HD Plasma) was to open an Explorer window and use the TV's auto adjust function. After it finished adjusting, the window displayed perfectly with no over-scanning. The PC is connected to the TV via the analogue PC input. It doesn't over-scan. It did before I adjusted it... well the picture was off to one side a little, but now it's perfect.
The TV is not an "early" model. It's a current model I purchased two weeks ago, and I updated it to the latest firmware a day later.

[hello_hello]

I'm pretty sure it's not your Bluray player that is doing the overscan but your TV. New TVs do overscanning as well because if they didn't you would obviously have situations in which you see nasty borders and blinking white lines at the edges (depending on the content but with regular TV programs you would see this sooner or later).

Thanks for the reply. Now the inevitable Ghitulescu nonsense has been dealt with.....

I understand over-scanning when it comes to playing DVD or free to air content, but when it comes to full HD stuff it does seem unnecessary. So far everything I've tried via the HDMI input (and I've tried pretty much everything except for an actual Bluray disc as I hardly ever play them) has been over-scanned, regardless of the original resolution.

I haven't found a setting to disable it yet, but I haven't as yet looked for one. I'll check it out. I've only owned the TV a couple of weeks so I haven't even taken the plastic wrapping off the manual yet. :)
I did disable the anti-burn in protection stuff, so I don't think it's causing the over-scanning. Is it possible the TV only over-scans when using some of the HDMI inputs but not others? I should read the manual.

Not that it matters too much anyway. It surprised me when comparing them but there's no doubt in my mind the PCs analogue input looks better than when playing the same video with the Sony player. It just looks a little more "natural". I thought I'd have to connect the PC via the DVI input for it to look as good, but now I won't bother, especially if it might end up over-scanned.
I'm quite happy continuing to use the PC as a playback device 99% of the time. It's virtually always on anyway, and I've been using a PC as my playback device for years.

[Ghitulescu]

... the first thing I did after connecting the PC to the TV (51" Full HD Plasma) was to open an Explorer window and use the TV's auto adjust function. After it finished adjusting, the window displayed perfectly with no over-scanning. The PC is connected to the TV via the analogue PC input. It doesn't over-scan. It did before I adjusted it... well the picture was off to one side a little, but now it's perfect.
The TV is not an "early" model. It's a current model I purchased two weeks ago, and I updated it to the latest firmware a day later.
Not that you want to be clever than you are in reality, but did you ever occur to you that the PC input of the TVs are specially designed for computers....... :) Of course the TV fits the image - otherwise, like for TV-Out outputs of various graphic cards, the START button and the status bar will be completely missing (unless one activates the overscan function of the drivers, that would add black borders to the standard windows image).
But, hey, we are again here at work explaining to hello_hello something that again doesn't anything have to do with the original topic.

[hello_hello]

No matter how quickly the latest Ghitulescu nonsense is dealt with, there's always more of it to come....

Not that you want to be clever than you are in reality, but did you ever occur to you that the PC input of the TVs are specially designed for computers.......

Well obviously it occurred to me, but obviously it didn't occur to you when you said "The overscan is always enabled (and cannot be disabled) on all analog inputs."

I assume even you are aware the PC video input is analogue? I did explain it fairly clearly for you in my last post.

I see you conveniently skipped over the part of my post which showed your statement "a FullHD TV doesn't do overscan at 1920x1080 via HDMI" was more of your usual nonsense..... instead picking off-topic irrelevancies to argue about in the hope nobody will notice.
Did you make up the HDMI over-scanning thing as you wrote your last post, or was it copied from a previously written piece of fiction?
__________________
It's not difficult to find the ideal signature. Difficult is not boring everyone to death with it.

[mpucoder]

Do I have to remind both of you about rule #4? Keep the conversation civil and on topic please.

[hello_hello]

Well, I discovered how to disable the HDMI overscanning our resident expert on everything claimed doesn't happen.

The TV has various screen adjustment settings but they're different according to the input being used at the time, which is maybe how I missed it. With the screen in normal 16:9 mode it overscans, but down past the various 4:3 and stretching options there's "fit to screen".

No more overscanning.... :)

PS. I also discovered (according to the manual) it disables overscanning when 1080i/p content is being inputted via the component video inputs. So much for Ghitulescu's the "overscanning always being enabled (and can't be disabled) on all analog inputs" misinformation.

[manono]

it's always the TV the device in the chain that does the overscan, not the player. A player does only upscaling/downscaling, and only when it's needed.
I know nothing about Blu-Ray players but many DVD players, and perhaps the majority of them, do crop the video before passing it on to the TV set. That amounts to the same thing as an overscan - missing picture.

And quite a few TV sets, even modern LCD HDTVs, also overscan by default. My Sony doesn't but in his most recent post hello_hello discovered how to turn off the overscan on his set.

And like mpucoder, I remind you two to keep it civil before this thread gets closed.

[Ghitulescu]

If you're right, manono, then there's a problem with the HQV DVD test disc :)
It can only work with "non-overscanning" DVD-players, otherwise the results will be absolutely catastrophic.
Any DVDplayer must obey the PAL standard, in Europe, and this includes the overscan. Philips (and Philips-based) players also have a setting to set the overscan, but only for left-right position (as not all TVs use 5-5%).

Maybe there's a distinction between NTSC (USA) and PAL (Europe) gear. I don't know.

And in Germany, which accounts for a good share of TV market, all the LCD models I've tested, all have the HDMI overscan set to OFF. Have this been done by the personnel, is the off-the-box setting, hard to say, but I think it's a Sisyphus work to install some 200 TVs just for fun.

[hello_hello]

Why do you test LCD TVs and how many models have you actually tested? 200? Where do you test them?

[manono]

It can only work with "non-overscanning" DVD-players, otherwise the results will be absolutely catastrophic.
I didn't say they overscan, but that they crop, many of them. Of course, that's not the same thing, but the effect is the same - missing picture for the viewer, in addition to whatever the TV set might overscan.

I wanted to link to The Secrets Of Home Theater DVD Player Shootout where you can see which players crop, but it seems to be down at the moment.

[Ghitulescu]

Why do you test LCD TVs and how many models have you actually tested? 200? Where do you test them?

In showrooms.

[hello_hello]

In showrooms.

LOL! I couldn't decide whether you must regularly go into showrooms and ask the sales guy if he'd mind you testing all the TV's on the showroom floor for their default overscan settings, or whether when you said "all the LCD models I've tested" meant 3 or 4 models (not 200 as you implied), or whether you were just making it up.

When I first powered up my TV it asked if I wanted to run it in "home mode" or "showroom mode". Are the default overscan settings any different between the two modes?

[Ghitulescu]

LOL! I couldn't decide whether you must regularly go into showrooms and ask the sales guy if he'd mind you testing all the TV's on the showroom floor for their default overscan settings, or whether when you said "all the LCD models I've tested" meant 3 or 4 models (not 200 as you implied), or whether you were just making it up.

When I first powered up my TV it asked if I wanted to run it in "home mode" or "showroom mode". Are the default overscan settings any different between the two modes?

No, that's the reason they are displayed for, to be tested. One is required not to change the wiring though ;)

The difference is in the colour saturation and luminosity.

[hello_hello]

The difference is in the colour saturation and luminosity.

So you know for a fact there's no difference in overscan settings?

I still can't imagine why anyone would wander around showrooms testing the default overscan setting on TVs. I'll try asking yet again....
As you seem to be speaking with authority on the default settings of all LCD TVs, how many have you actually tested?

[jmac698]

Interesting thing is, I just tested the actual aspect ratio of dvd's, by simply comparing them to the HD version. I found that a vertical zoom of 1.02 from hd to sd, was the only change
HD.resize(720,480).zoom(1,1.02)
If you can figure that out, you've got the aspect ratio...
My point was, regardless of any standards out there, they are sometimes ignored. The aspect ratio only depends on the editing program used to finalize the dvd. There's only a few major programs, they have certain defaults, depending on version. The example I just gave certainly doesn't fit any standard.

As to why there's pillarboxing, I've seen this sometimes and it corresponded fairly well to the 711 pixel ratio, so my guess is that it was digitized from an analog tape. Modern productions from digital intermediate or pure digital wouldn't naturally have this. And yes, I saw that article at hometheatrehifi.com and some dvd players crop. Blurray players also mess with the gamma! I know my lcd tv has a weird aspect ratio. I've done some work calibrating different capture cards and playback devices too, don't know off hand though.

[SeeMoreDigital]

I thought I'd ask here instead and give you the opportunity to put your foot in your mouth.Oh jeez...

That kind of retort seems very much like something yetanotherid (http://forum.doom9.org/member.php?u=148781) would have said :eek:

[hello_hello]

Oh jeez...

That kind of retort seems very much like something yetanotherid (http://forum.doom9.org/member.php?u=148781) would have said

Possibly, as it seems an appropriate response to a sarcastic reply containing misinformation. What's your point?

[setarip_old]

@hello_hello Possibly, as it seems an appropriate response to a sarcastic reply containing misinformation. What's your point?
The point is, that a more appropriate course of action would be to privately advise a moderator of what you perceive to be an inappropriate response by another member, rather than the two of you continuing to clutter up a multitude of threads here with nonsensical arguments, sarcasm, and flaming - unless, of course, you're concerned tha the moderators may realize what both SeeMoreDigital (today) and I (several months ago) have each alluded to as a distinct possibility regarding the screenname "yetanotherid"...

[Ghitulescu]

@hello_hello
The point is, that a more appropriate course of action would be to privately advise a moderator of what you perceive to be an inappropriate response by another member, rather than the two of you continuing to clutter up a multitude of threads here with nonsensical arguments, sarcasm, and flaming - unless, of course, you're concerned tha the moderators may realize what both SeeMoreDigital (today) and I (several months ago) have each alluded to as a distinct possibility regarding the screenname "yetanotherid"...

Strange that you two realized this only now :) after I've said it long time ago (and the proof of this equality is on file).

[Ghitulescu]

Interesting thing is, I just tested the actual aspect ratio of dvd's, by simply comparing them to the HD version. I found that a vertical zoom of 1.02 from hd to sd, was the only change
HD.resize(720,480).zoom(1,1.02)
If you can figure that out, you've got the aspect ratio...
My point was, regardless of any standards out there, they are sometimes ignored. The aspect ratio only depends on the editing program used to finalize the dvd. There's only a few major programs, they have certain defaults, depending on version. The example I just gave certainly doesn't fit any standard.

It's not that simple. Even assuming the material for both the DVD and the BD is the same (as a film roll), each studio crops the workprint as they want before even feeding it to an encoder, because the requirements for a DVD are different than those for a BD. There are even versions of the same movie on the same medium that are differently cropped (see Ben Hur 1959). To simplify the calculations, non-ITU resizing for DVDs is a good approximation.
In the end, setting aside the tiny distortion caused by a wrong DAR is a job for purists. Average Joe won't care. He didn't care nor noticed before (the national sport of "backing up my collection to AVI" is quite old now, but the "correct DAR" issue appeared much later).

[hello_hello]

Strange that you two realized this only now :) after I've said it long time ago (and the proof of this equality is on file).

Yeah, there's an obvious equality when it comes to the common sense questioning of bad advice.
Was it supposed to be a secret? I don't use the old ID any more. I haven't used it since I started using this one. Should I see if I can delete the old one?

I thought starting again with a new ID might help "keep the peace" a little but it backfired on me. It just turned out to be an addition to your arsenal of excuses for avoiding the topic being discussed.

[hello_hello]

The point is, that a more appropriate course of action would be to privately advise a moderator of what you perceive to be an inappropriate response by another member, rather than the two of you continuing to clutter up a multitude of threads here with nonsensical arguments, sarcasm, and flaming - unless, of course, you're concerned tha the moderators may realize what both SeeMoreDigital (today) and I (several months ago) have each alluded to as a distinct possibility regarding the screenname "yetanotherid"... [/Color]

Feel free to use a more of appropriate course of action yourself instead of constantly cluttering up threads with complaints regarding unrelated conversations in different threads.

The only reason my discussions clutter up threads is because I've got to repeat the same questions over and over in a vain attempt to get our resident expert on everything to provide answers on a topic on which he professes to be the expert.

How many posts have I wasted so far in this thread simply trying to get a simple answer to the question "how many TVs have you tested", which is a fairly valid question given he seems confident he can offer advice as to the way "all" TVs are set up?
Just another small example of why we have to keep going round and round in circles while Ghitulescu tries to hide the fact he pulled his latest facts out of a hat.

My old ID hasn't been banned, I just don't use it any more. If I could have worked out how to "unregister" it, I would have done so. If you want to run to the mods about it, feel free to ask them to delete it.

[Ghitulescu]

Yeah, there's an obvious equality when it comes to the common sense questioning of bad advice.
Was it supposed to be a secret? I don't use the old ID any more. I haven't used it since I started using this one. Should I see if I can delete the old one?

I thought starting again with a new ID might help "keep the peace" a little but it backfired on me. It just turned out to be an addition to your arsenal of excuses for avoiding the topic being discussed.

You did this to make me believe I am wrong in my assertions, when allegedly two "different" people contest my arguments. A secondary goal was to appear with a new immaculate profile, as yetanotherid was already compromised. As the account before him.

But you can't "make peace" without changing your habits. You polished however your spam-technique managing to get less strikes for the same offences.

I answer your questions in a topic of yours. Otherwise it might be an off-topic discussion. Like this one.

[hello_hello]

You did this to make me believe I am wrong in my assertions, when allegedly two "different" people contest my arguments. A secondary goal was to appear with a new immaculate profile, as yetanotherid was already compromised. As the account before him.

What a load of waffle. I thought it might stop you crying wolf for a bit.

As the account before him.

Proof positive you simply make stuff up. I have never used any other IDs but this one and my previous one.

You polished however your spam-technique managing to get less strikes for the same offences.

Another example of you simply making stuff up. How on earth could you possibly know how many strikes someone else has had?

I answer your questions in a topic of yours. Otherwise it might be an off-topic discussion. Like this one.

There's an excuse laced with self-admitted hypocrisy. You're happy to clutter threads by crying wolf and complaining about motives for posting, but when you make a claim based on "all the TVs you've tested" and you're asked how many you've tested, the only thing off-topic are your excuses for not answering.

So are you going to answer it? You made a claim regarding the way "all TVs" work. To know whether the claim is valid, it's not unreasonable to show how large a sample you've personally tested? I've only tested one, but I can tell you you're generalisation was wrong when it comes to the way it works.

Oh.... and the question about the difference in over-scan settings when a TV's running in "showroom mode".... you're still ignoring that one too.

No matter how quickly the latest Ghitulescu nonsense is dealt with, there's always more of it to come....

[Ghitulescu]

Oh.... and the question about the difference in over-scan settings when a TV's running in "showroom mode".... you're still ignoring that one too.

Not only that you can't read, the answer being in #20 (http://forum.doom9.org/showpost.php?p=1524501&postcount=20), but also insist on off-topic issues.

The topic is why do have DVDs sometimes (not always -> http://forum.doom9.org/showthread.php?p=1494312#post1494312) black bars, and it has been answered.

The BBC DVDs from TV shows I have also have black borders around the image.

[hello_hello]

Not only that you can't read, the answer being in #20 (http://forum.doom9.org/showpost.php?p=1524501&postcount=20), but also insist on off-topic issues.

I can read, I didn't ask about saturation and luminosity. That's obvious.
Did you miss my follow up post where I asked you to confirm there was no differenced in overscan settings?
Did you miss where I asked how many TVs you've tested, or do you have trouble reading? :)

Directly from the Ghitulescu handbook of excuses for not answering questions.
Excuse #436: Pretend a renewed interest in the original post. If the question has long been asked and answered, pretend it hasn't.

I don't think the OP cares about your BBC DVDs. He did however want to know why some have black borders.

[hello_hello]

Interesting thing is, I just tested the actual aspect ratio of dvd's, by simply comparing them to the HD version. I found that a vertical zoom of 1.02 from hd to sd, was the only change
HD.resize(720,480).zoom(1,1.02)
If you can figure that out, you've got the aspect ratio...

How were you comparing them to the DVDs? Were you playing the original DVDs with a software media player or were you comparing DVD encodes with the HD versions etc?

I've compared lots of DVDs to their HD counterparts, and while they are sometimes cropped differently (which isn't really relevant to the aspect ratio used), I've found 99% of the time a non-ITU resized DVD will display with the same aspect ratio as the HD version. If not 100% exactly, still with nowhere near the difference you're describing.

A 1.02% vertical zoom requirement (I assume you mean you've got to vertically zoom the DVD version?) would seem to indicate you're comparing a DVD which has wrongly been resized using the ITU resize method with a HD equivalent.

The HD version you're using for testing.... is it the original video from the disc or an encode? I'm just trying to eliminate the possibility you're comparing the DVD to a HD encode which had it's aspect ratio distorted a little when it was encoded.

[jmac698]

I almost stopped reading when I found your question. I wish someone would just delete the off-topic posts and be done with it.

I'm comparing the raw video files. The code shows I resized the HD to match the SD, the SD is cropped relatively to the HD (the HD was zoomed 1.02 times). And this is all relative to a straight 720x480=16:9. So that shows that the versions are *nearly* assuming 720x480=16:9. Measuring the aspect of different playback devices is a completely different story, however I believe that analog SD outputs are going to be ITU based because it's the easiest to do electronically. My TV zooms everything a bit in analog (yep overscan on an LCD), and my players look different in analog or HDMI outputs, so it's just a mess.

And while there are differenly cropped versions of movies, that has nothing to do with the aspect, which I assume can always be measured by comparing against the HD version, which I assume is always 1:1.

[hello_hello]

I almost stopped reading when I found your question. I wish someone would just delete the off-topic posts and be done with it.

I actually posted it a couple of times, deleted it each time, then posted it again before I finally managed to stop it being surrounded by the off-topic nonsense.

I'm comparing the raw video files. The code shows I resized the HD to match the SD, the SD is cropped relatively to the HD (the HD was zoomed 1.02 times). And this is all relative to a straight 720x480=16:9. So that shows that the versions are *nearly* assuming 720x480=16:9.

Maybe I'm being dumb, but I'm still not 100% sure I'm following your method. The way I'm understanding it, you resized the HD video to 720x480 in order to compare it to the DVD version and found the resized HD version had to be vertically zoomed 1.02 times to match the DVD??
I'm probably still not following you properly, but when you resized the HD version are you sure there was no aspect ratio distortion introduced in the process?

The way I've compared the two.... well to be honest I've not very often compared the HD version to the DVD directly, but I have quite often compared the HD version to older AVI/DVD encodes. I always use AutoGK for encoding AVIs and I know it doesn't distort the aspect ratio in the process because I have compared many AVI encodes with the original DVD and I know the (picture) aspect ratio is always kept intact (because of the way it crops and resizes).

So in my case I open the AVI/DVD encode with MPC-HC (which was invariably created using the non-ITU resize method and will have exactly the same aspect ratio as the original DVD) and the open the HD version with a second instance of MPC-HC. I then run both versions in a maximised window and compare the two for aspect ratio differences.
Differences in cropping aside, I've found in by far the majority of cases, the old AVI encode and the HD version have virtually identical aspect ratios. That's why when you said you needed to zoom the (resized) HD version vertically by 1.02 times to get the two versions to match, I found it quite strange.

I'm still not sure I understand why you're resizing the HD version to 720x480 in order to compare it with the DVD version, or how you're calculating the resizing. Wouldn't it be simpler to open the HD version in one instance of a player, open the DVD version in a second instance, maximise them both and then compare the aspect ratios? If you did that and the aspect ratios matched, it'd mean the DVD uses straight 16:9 resizing and not ITU resizing, as most software players just use the straight 16:9 resizing method.

And while there are differenly cropped versions of movies, that has nothing to do with the aspect, which I assume can always be measured by comparing against the HD version, which I assume is always 1:1.

Yes, the cropping "information" was pretty much irrelevant.

[lovelove]

Hi. I was wondering why many DVDs have a 16px black frame left+right (but not on top/bottom) ?

screenshot:
http://img847.imageshack.us/img847/4984/vlcsnap2011081111h42m37.jpg


Although interesting, the discussion here does not really address the core question, namely why anyone would choose to select an image area without data (black) for DVD mastering? If the original footage AR does not exactly match the DVD AR, it could be fitted outside instead of inside, i.e. choosing the height so that the width fits and no black space needs to be added.

[jmac698]

it could be fitted outside instead of inside

That was my point as well, you can fill that space, but to do so would involve a little bit of cropping. The 18 extra pixels should be wasted on TV, theoretically, and never displayed.


compare in mpc...

I'm taking a half-step to that - I eliminate the player from the equation, and simply compare the actual raw video, without any aspect correction. What is the exact formula to match HD to SD? From that I can derive the actual aspect ratio of the SD version. If that aspect happens to match ITU, then great, mpc/player x will play it in proper aspect. And you understood my matching, the code is explicit:

SD=HD.bilinearresize(720,480).zoom(1,1.02)

I haven't calculated what PAR that gives me in the end, just off the top of my head I don't think it's ITU. And with your method, how do you know that MPC isn't in error? Do you know for sure it's following ITU? In the end, all that matters to *you* is that they play in proper aspect on the player/TV which you use most often, if you really want to check that it's proper (and if not, a quick pass with mpeg4modifier can fix it).


Also I plan to do a better measurement than eyeballing it; I was going to use mvtools to match images to the nearest pixel.

[hello_hello]

Although interesting, the discussion here does not really address the core question, namely why anyone would choose to select an image area without data (black) for DVD mastering?

Analog-to-digital conversion (http://en.wikipedia.org/wiki/Pixel_aspect_ratio#Analog-to-digital_conversion)

[hello_hello]

I'm taking a half-step to that - I eliminate the player from the equation, and simply compare the actual raw video, without any aspect correction. What is the exact formula to match HD to SD? From that I can derive the actual aspect ratio of the SD version. If that aspect happens to match ITU, then great, mpc/player x will play it in proper aspect.

MPC, and to the best of my knowledge all other software media players, don't use ITU resizing. They all use straight 16:9 or 4:3 resizing.

And you understood my matching, the code is explicit:

SD=HD.bilinearresize(720,480).zoom(1,1.02)

I haven't calculated what PAR that gives me in the end, just off the top of my head I don't think it's ITU.

I do understand what you're getting at (I think).
I guess in theory if the two videos were identical, and if you resized the HD version to 720x480, without aspect ratio correction the resized HD version and the DVD version should look identical. But you're saying you had to zoom the height of the resized HD version in order to get them to match.

I probably shouldn't be thinking about this while I'm suffering from a lack of sleep, but.....
Looking at it from the opposite direction.... although I can't get it straight in my head for some reason.... the 1.02% zoom you describe happens to be pretty much the difference in PAR between ITU and non-ITU resizing. You've had to resize while including a 1.02% zoom in order to match the DVD, therefore, wouldn't it follow in order for the DVD to match the HD video's aspect ratio it must use ITU resizing?
Had you picked a different DVD, one which doesn't use ITU resizing, the 1.02% zoom wouldn't have been required when resizing the HD version because the DVD would also use straight 16:9 resizing.

And with your method, how do you know that MPC isn't in error? Do you know for sure it's following ITU?

I just checked some vob files on my PC against the non-ITU resized encodes using MPC. They display with the exact same aspect ratio so I'm 100% sure MPC doesn't use ITU resizing.
Knowing how MPC displays DVDs, for me it's easier to use MPC to compare the aspect ratio of the DVD with the HD version. If the aspect ratios match then the DVD uses straight 16:9 resizing, if the DVD looks a little vertically stretched compared to the HD version then the DVD should be resized using the ITU method.
Your way of working it out seems to require too much reverse engineering (thinking) for me. At least at the moment. Maybe when I'm more awake...

[jmac698]

Yes your method is simple, I agree, but it's only valid for that combination of player and file, which may be fine if that's what you mostly use for playback. And it seems you agree that dvd's are often sized as straight 16:9. I can't explain the small discrepancy I've found, but the amount is too small to be ITU. And I, too, could answer this better if I thought about it more ;)

[hello_hello]

I'm not sure I understand your logic when you say my method is only valid for my combination of player and file. If I know how the player resizes a DVD then it's resizing is valid for every file. Generally when using a GUI for encoding you have to pick one resize method or the other. If the DVD's aspect ratio happens to vary a little from one of those resize methods it's not going to be by much, and not by enough I'd try to work it out.

The ITU storage aspect ratio for a 16:9 NTSC DVD is 1.823169. A straight resize is 1.777778. The difference is about 1.0255%

Edit: The above should say "The ITU Display aspect ratio for a 16:9 NTSC DVD is 1.823169. A straight resize is 1.777778. The difference is about 1.0255%"

A more accurate way to compare them (rather than zoom) might be to take a screen shot of each version (MPC saves screen shots using the video's storage aspect ratio). Then opening the image with a program such as Irfanview, you could resize and crop the DVD screen shot to an exact number of pixels, either in width or height.

[jmac698]

aha! Those are the numbers I was looking for, so then it does make sense that my test was ITU but by cropping the source, rather than adding borders.
And I say by file because, there's been issues reported where the aspect was set to one thing in the container and another thing in the video (pertaining to MKV, mp4).
And I was comparing not with snapshots, but directly in avsmod, which I know to be accurate.
Example:

I set the display width/height of video track in mp4 to 0/0 and hope the player always ignore this and take width/height*sar in bitstream. Unfortunately QT fails to do so.

http://forum.doom9.org/showthread.php?t=152419

[hello_hello]

In my limited experience after making a mess of MKV aspect ratios at one stage I found MPC-HC seems to follow the aspect ratio in the video stream when using it's internal filters, and when using the Haali media splitter it follows the container aspect ratio. My Sony Bluray player follows the container aspect ratio.
I've not installed QT on a PC for years so I don't know what it does.

I actually messed up the container aspect ratios when using the tool you linked to in order to fix the colorimetry of some x264 anamorphic DVD encodes. I didn't realise it at the time, but the remuxing was stripping out the container aspect ratio, so MediaInfo was reporting the "Display Aspect Ratio" as being the same as the storage aspect ratio, and the "Original Display Aspect Ratio" as being the aspect ratio in the video stream. When I later used MKVToolnix to repair the container aspect ratio, MediaInfo no longer reported any "Original Aspect ratio" at all, just the correct "Display Aspect Ratio".
None of my other anamorphic encodes had a different "DAR" and "Original DAR", and I checked quite a few of them at the time.

Irfanview should be good for comparing screenshots. If you save a 720x480 frame using MPC-HC and open the image with Irfanview, under Image/Information Irfanview will tell you the image has a resolution of 720x480 pixels. I've used Irfanview a few times to resize images when trying to decide whether "to ITU or not to ITU" as you can resize to a predetermined width and height in pixels, or by a percentage.

[Ghitulescu]

I'm not sure I understand your logic when you say my method is only valid for my combination of player and file. If I know how the player resizes a DVD then it's resizing is valid for every file.

Nope, resizing a DVD has to fulfil some strict requirements of a standard. Resizing an AVI/MKV is more or less relaxed, as no legal constraints are to be fulfilled, especially when no valid frame sizes are found like 628 x 234. It also depends on how primary output is internally set (HDMI or SCART).

[hello_hello]

Sigh.....
Let me explain. In the context of the post you quoted...
"DVD" = "File".
"Player" = "MPC-HC".
We were only referring to comparing the DVD and Bluray versions of video, and the way MPC-HC resizes DVDs so it can be used to compare DVD/Bluray aspect ratios.
Your AVI/MKV resizing comments, even if they're co-incidentally accurate, aren't really relevant to the discussion.

Nope, resizing a DVD has to fulfil some strict requirements of a standard.

ITU is just a recommendation. It's not mandatory. It's really at the discretion of the HW manufacturer whether to comply or not with the ITU-R provisions and how to do it.

So resizing a DVD has to fulfil some strict requirements of a standard which is just a recommendation and at the discretion of the manufacturer whether they comply.

Sigh......

[Ghitulescu]

Sigh.....
Let me explain. In the context of the post you quoted...
"DVD" = "File".
"Player" = "MPC-HC".
We were only referring to comparing the DVD and Bluray versions of video, and the way MPC-HC resizes DVDs so it can be used to compare DVD/Bluray aspect ratios.
Your AVI/MKV resizing comments aren't really relevant to the discussion.
Unfortunately you were specifically discussing about MKV, when jmac698 questioned your logic. And the post I quoted was discussing MKV and file in the same context.
If you're discussing DVD all this time when you wrote MKV and files, there's no wonder jmac698 questioned this logic.

And still there's a difference between MPC-HC and a real (hardware) player, which make your comparison a bit forced.
So resizing a DVD has to fulfil some strict requirements of a standard which is just a recommendation and at the discretion of the manufacturer whether they comply.

Sigh......
Thank you for another sample of your knowledge. And gratuitous malice.
Try SMPTE :) for a mandatory standard. There are several standards for both analog and digital (scanlines, colorimetry, sampling, interfaces etc.), not just that recommendation.

[jmac698]

@hello

Originally Posted by jmac698 View Post
Thank you for another sample of your knowledge. And gratuitous malice.

I don't know how you extracted this quote, but it makes it appear as if I said that, when in fact it should be attributed to Ghitulescu.
Anyhow, I think I'll finish my tests so this discussion will end, though to me it added a bit of humour to my day :) (as a non-involved observer, no offense).

[hello_hello]

jmac698,
My mistake and apologies. I've fixed the quote. Actually, I deleted the whole post. Apparently I thought I was referring to the resizing of DVDs using MPC-HC in the post Ghitulescu quoted, and the term "MKV" hadn't even been used up until that point, but he knows better so we'll leave it at that.

So rather than let the arguing for the sake of arguing get in the way, how did your tests go? Did you come to any further conclusions or have you settled on either ITU or non-ITU?

[jmac698]

Ok, I have a precise answer.

Method of Measurement
Instead of trying to do this in script, I just used http://www.fiji.sc Load two images, then go plugins, feature extraction, extract SIFT correspondences. What this does is find similar spots in each image, then collectively finds the best stretching that fits. The result is in the form of a Java AffineTransform function. For my example 1 I got:

AffineTransform[[1.000550508499146, -0.002122041769326, -0.329149395227432],
[7.0812115155E-5, 0.98687869310379, 5.01885986328125]])

You need to know what these numbers refer to, so you can plug them into this matrix:

[ x'] [ m00 m01 m02 ] [ x ] [ m00x + m01y + m02 ]
[ y'] = [ m10 m11 m12 ] [ y ] = [ m10x + m11y + m12 ]
[ 1 ] [ 0 0 1 ] [ 1 ] [ 1 ]

I don't expect any rotation, so I'll just ignore that aspect (a rotation is done by adding a bit of Y to X, for example). So ignore any numbers very close to 1 or 0. What's left is 0.98687869310379 which is the stretch in Y to create 1:1 aspect. The 5 is how much the image has to be shifted in Y, I'd guess it has to be moved down.
The result itself is accurate within .5 pixel, so it's a good solution (my resizers would probably have an effect here).

The Measured PAR
So my final PAR is (1/.9868*16/9)=1.8. You gave me 1.82 for ITU which is quite close. Note that squishing the DVD to 99% vertically is the same as stretching it 1.01% horizontally. So the best playback for this video is to stretch to 864.7x480.

Derivation of ITU 16:9 PAR
I'm going to derive that again for myself:
1:1 analog is 710.85x486, or 702.07x480. So I want to know what 720 would give. 720/702.07*16/9=1.823. We agree!

Is it ITU?
The .02 difference may not seem like a lot, but we have to compute the significance of this. My error measure of .5 was the maximum error in the selected corresponding pixels (I'm guessing), but I don't know which pixel it was, I would like to use pixels in the 4 corners for best accuracy. I also compared both images resized to 854x480. It should be 853.333. So let's say the max error is .66+.5 or 854.5/853.33 or 1.001. My actual error to ITU is 1.01, so no I'm not willing to say it's exactly ITU.