Correct aspect ratio of 4:3 DVD

[Slogra]

I'm ripping PAL DVD's without resizing. So resolution is 720x576.

Now I want to set the aspect ratio in the video properties using MKVToolNix. However 4:3 is probably not correct due to overscan. I think should be a bit wider.

What aspect ratio should i use?

(this question probably has been answered many times before, but i could not find the answer)

[SeeMoreDigital]

Originally Posted by Slogra

I'm ripping PAL DVD's without resizing. So resolution is 720x576.

Now I want to set the aspect ratio in the video properties using MKVToolNix. However 4:3 is probably not correct due to overscan. I think should be a bit wider.

What aspect ratio should i use?

(this question probably has been answered many times before, but i could not find the answer)

Can you take a still image capture of your source and upload it to a sharing site?

[Slogra]

I meant in general, not specifically to this DVD.
I vaguely remember that the the image should be slightly more stretched horizontally.


I think old programs like guardian knot did do something like that. So I've just installed Gordian Knot and found what i was looking for at options:

"Follow ITU-R BT.601 Standard
It's recommended to leave this checked.
It is the correct resizing method for DVD.
If you uncheck 'ITU-Standard' the resizing method will be similar to the regular
one used in software DVD-players like WinDVD and PowerDVD and described in most guides.

But actually this will result in an aspect ratio that is ~2.5% off!
This fact was pointed out to me by jackei, the author of dvd2avi."


I still don't know what the aspect ratio should be though.

[hello_hello]

There's a list of pixel aspect ratios here:
https://forum.doom9.org/showthread.p...27#post1058927

Gordian Knot probably uses the "almost exact ITU" PARs.

If I remember correctly MKVToolNix doesn't store a pixel aspect ratio as such, but it calculates the display dimensions and stores them instead, rounded to whole pixel dimensions. You can only judge what looks correct to you by comparing the resizing based on each PAR, although it's hard to do because if you look at an object that should be round (for example), based on experience your brain will adjust so it looks round, assuming it's only a little stretched or squished, so if you switch to looking at the same frame with a different PAR, it can look slightly wrong at first even if it isn't. The closest you can get to determining the correct aspect ratio is to find a straight-on shot of something square or round, taking a screenshot and drawing a circle or square around it with an image editing program.

For 4:3 DVDs I always use the MPEG-4 PARs. They result in a display aspect ratio of 15:11 for both PAL and NTSC DVDs (16:9 DVDs tend to use the generic PARs unless they're old). The mpeg4 PARs are the digital equivalent of the analogue ITU PARs and close enough to being the same not to matter, but the numbers are much easier to remember.

720 x (12 / 11) = 785.4545
720 x (12 / 11) / 576 = 1.3636 or 15:11

If you're an Avisynth user, this function makes it easy to compare sample/pixel aspect ratios by opening a video and specifying either a display or sample aspect ratio, then specifying something different and refreshing the script.
CropResize(InSAR=12.0/11.0, ResizeWO=true) would simply resize the width based on the display or sample aspect ratio you specify. It resizes to mod4 dimensions by default but it'll crop a pixel or two from the width if necessary in order to resize without distorting the picture.

If you don't want to go to all that trouble, the MPEG-4 (or ITU) PARs would be my default choice for 4:3 DVDs.

[Slogra]

Thank you for you detailed answer.

Yes, 15:11 is what i was looking for. In MKVToolNix you can just input 15/11 as aspect ratio. I don't know how it actually stores that data.

When i play the video using MPC-HC and make a screenshot. The screenshot is 785x576.
Whereas 4:3 the screenshot is 768x576.

I'm reripping some old DVD to a better format and method. Currently i'm ripping Blackadder 3. I believe the BBC respected the standards for their DVDs. So i believe 15:11 should be right aspect ration.

From a screenshot it's hard to tell. I think Rowan Atkinson should have a fairy round head, so then 15:11 is the way to go.

[hello_hello]

For both PAL and NTSC the mpeg4 PARs would mean 704 pixels of the width result in an exact 4:3 image, and even though I've never seen one, 704x576 is a DVD compliant resolution for 4:3 (but not 16:9).

704 x (12 / 11) = 768
704 x (12 / 11) / 576 = 1.3333 or 4:3

So if you crop exactly 8 pixels from each side, which is often only black, you're left with 4:3.

Often it's rounded to 704 anyway, but for PAL ITU aspect ratios I recall reading it's actually 702 pixels rather than 704, making the total DAR just a tiny bit wider, although when I did the math the ITU PAR gave me 703.22 pixels, and it's the "almost exact" ITU PAR from the list that works out to exactly 702, but a quick search turned up two PDFs containing BBC broadcast standards and both refer to 702 pixels.
https://duckduckgo.com/?q=bbc+video+...l+specs&ia=web
There's also another PDF here that says the same:
BBC - Commissioning - A Guide to Picture Size - Avid Technology

So for anything that started life as analogue video, 15:11 (or 20:11) would be a safe bet, but I'm fairly sure all the 4:3 DVDs I've met have had an ITU/mpeg4 PAR regardless of the source, and I tend to assume 20:11 is the correct DAR for any 16:9 DVD from the BBC, although I'm sure most other 16:9 DVDs are 16:9 unless they're old and have a lot of black each side.

It's just a guess as to whether the DVDs were created using the original ITU PARs though, or even if the video was transferred correctly. I was involved in a thread here a while back (a search would probably find it) discussing the Bluray release of Faulty Towers, and I'm fairly sure the original video was upscaled based on the assumption it's exactly 4:3, so I'm quite sure the aspect ratio of the Bluray release is a little squished, and it definitely is when compared to the DVD version if you assume the DVD is 15:11.

[Slogra]

Thanks again for all the info.
I'll use 15:11.
If i ever change my mind, i can easily change mkv aspect ratio "tag" to something else.

[Emulgator]

RA's left eye, buttons, woodwork cutouts: I would tend to the narrower screenshot being correct.
The black bars just pad the active picture sample width (702) up to 720 samples.
720 total D1 sample width minus 702 analog-picture-content-sampled-@13,5MHz-width would leave 18 samples per line without content.
You would pad with black bar widths of maybe 10 left plus 8 right.
For the 767 wide screenshot I get 14 width left + 10 width right = 24 width of black bars.
Divide that by 767/720 and I get a sampled black bar width of ~22 samples.
Maybe 2 more samples left and 2 more samples right had been cropped to get rid of line lead-in and lead-out ?
Now everything fits.

[Sharc]

The wider picture is correct IMO. The inner (active) picture (excluding the side bars) is pretty close to 4:3, as it's supposed to be for a sampling rate of 13.5MHz.
The overall picture including the bars is 15:11 (1.36). Eye and buttons ar too small and not exactly perpendicular to be used for the "circle test", I think.

The second picture is horizontally squashed IMO.

[SeeMoreDigital]

As requested before: "Can you take a still image capture of your source and upload it to a sharing site?"

[StainlessS]

The more narrow image looks more like him to me, the upper image being (head) too wide.
(However, given that its Rowan Atkinson, could be either, he's a peculiar mix of distortion)

[hello_hello]

The BBC pdf I linked to earlier seems to include some rounding when showing the math, but it explains how to reverse engineer a 4:3 DVD from a square pixel image this way:
BBC - Commissioning - A Guide to Picture Size - Avid Technology

When making a 4:3 graphic on a square pixel device for conversion to a 720 by 576 (non square pixel) video image,
the width of the 768 by 576 (square pixel) image must be increased to: 768 + 20 = 788 square pixels making the 4:3 the square pixel graphic 788 by 576.

The first image over-shoots the 20 pixels mentioned in the pdf for a total of 24 (or maybe 25). I resized the width to 788 before counting because the image is only 784 pixels wide.

For the record the BBC also says a 16:9 image would need 26 pixels of black in total.

When making a 16:9 graphic on a square pixel device for conversion to a 720 by 576 widescreen video picture,
the width of the 1024 by 576 image must be increased: 26 +1024 = 1050 by 576 square pixels

Of course this source started out as analogue video rather than a square pixel image, and sometimes there's a huge variation in the amount of black, even from scene to scene, for reasons I don't understand. None of the BBC info proves a DVD was created without any aspect ratio fudging or using the correct PAR anyway, although in this case my vote for the least amount of incorrectness would go to the top image.

[flossy_cake]

Yes, the aspect ratio of DVDs is often wrong by +/-2.272% in width. From what I can tell this issue may originate from something like analogue video signals being 704 wide whereas digital is 720 wide. 720/704 = 1.02272 so that's your width multiplier if the DVD looks too skinny. If faces look too fat then it's going to be the reverse 704/720 = 0.9777 or 97.7% width.

How you implement it is going to affect things as well. I use Avisynth which works in square pixels so I already know 768x576 is 4:3 square pixels, so I just multiply the 768 by either 1.02272 or 0.9777.

Some DVDs have a mix of +/-2% widths, like Skippy the Bush Kangaroo PAL DVD has correct intro sequence but 2% wide faces for the show itself. Then a few episodes later it seems to come good, but I'm not even 100% sure how their faces are supposed to look. It's an annoying issue that we are saddled with and I try not to stress over it too much.

[flossy_cake]

Originally Posted by flossy_cake

720/704 = 1.02272 so that's your width multiplier if the DVD looks too skinny. If faces look too fat then it's going to be the reverse 704/720 = 0.9777 or 97.7% width....I already know 768x576 is 4:3 square pixels, so I just multiply the 768 by either 1.02272 or 0.9777.

Now that I think about it, it seems wrong to apply a nonsquare pixel multiplier to square pixels, doesn't it?

On the other hand, the resulting change in DAR should be the same, shouldn't it? Like, if we multiply nonsquare pixel width by 2x and get 1440 wide, that will still make the resulting DAR twice as wide, which is the same as what happens when multiplying the square pixel width by 2x as well.

[Sharc]

Originally Posted by flossy_cake

Now that I think about it, it seems wrong to apply a nonsquare pixel multiplier to square pixels, doesn't it?

On the other hand, the resulting change in DAR should be the same, shouldn't it? Like, if we multiply nonsquare pixel width by 2x and get 1440 wide, that will still make the resulting DAR twice as wide, which is the same as what happens when multiplying the square pixel width by 2x as well.

Yes. Multiplying affects both the picture's dimension and the Pixel Aspect Ratio.
Imagine a piece of (elastic) paper with squares (representing pixels with a Pixel Aspect Ratio 1:1) or rectangles (representing some non-square Pixel Aspect Ratio like on DVD) on it. When you multiply (means stretch or shrink) the "paper" in one dimension (width or hight) you change both the overall size of the sheet and the shape of the squares or rectangles on it.
On the other hand, Cropping (=changing the outer dimensions of the paper with a scissors) only changes the overall dimensions of the sheet, leaving the squares or rectangles (representing the pixels) on it intact. Hence Cropping does not affect the Pixel Aspect Ratio.
Zooming (=multiplying width and hight by the same factor) keeps both the width/high ratio of the sheet and the width/hight ratio of the squares or rectangles on it intact.

[flossy_cake]

This is from the 1999 Blackadder movie, scaled to 1024x576 square pixels



Either his head got wider with age or its the focal length



In light of this I'm not really sure we can just eyeball which one is "correct" .