Aspect Ratio

[ArielHatotah]

I didn't get something about aspect ratio.
Supposing I'm playing back on a widescreen TV.
In DVDs, when written Anamorphic Widescreen, it means that the picture, supposed originally filmed at 16:9 aspect ratio, is stored as "squeezed" picture (the horizontal resolution is squeezed), thus fulfill the whole frame (suppose 720*480 (NTSC)). Therefore, when playing back, the picture is horizontally stretched back.
Till now I understood. I hope it's true.
But if written on the DVD "Letterboxed Widescreen" with a DAR of 16:9, then what is going on with the picture? How is the picture stored on the DVD?
Supposing the same original picture (filmed at 16:9 aspect ratio), is the picture stored with black bars up and down (that's the meaning of "letterboxed")? But if it is, then when playing back, the picture is stretched (because of the DAR), and then the picture should be distorted. Isn't it? Am I wrong?
Of course I'm wrong, because it can't be that the picture will be distorted. So what is my mistake? Where am I wrong?

[foxyshadis]

Three snaps I whipped up that might help:
4:3 (http://foxyshadis.slightlydark.com/random/ar1.33.jpg)
16:9 (http://foxyshadis.slightlydark.com/random/ar1.77.jpg)
2.35 (http://foxyshadis.slightlydark.com/random/ar2.35.jpg)
Of course there's a lot more involved with active areas and overscan, but I don't personally care about any of that.

You're correct that mattes are used, and then depending on the output signal, more mattes may be added by the player, or mattes may be removed to fill more of the screen.

[ArielHatotah]

O.K. Thanks.
But now supposing the same film, with a picture filmed at 16:9 aspect ratio (like the second example that foxyshadis gave).
And one want to encode it on a DVD with Anamorphic Widescreen, and another want to encode it on a DVD with Letterboxed Widescreen. Both with a DAR of 16:9. What will be the difference between the two?

[SeeMoreDigital]

Have a look here for PAL: -

http://www.seemoredigital.net/03_Video_Info.html

and here for NTSC: -

http://www.seemoredigital.net/03_Video_Info.html

[ArielHatotah]

Thank you, SeeMoreDigital. Very nice site.
So if I understood, when using Anamorphic Widescreen, the actual resolution is 853*480 (in NTSC), and it is stored on the DVD as 720*480.
Then if I convert it to MPEG4 (AVI), I can use the resolution 848*480 for example. But I know that if you convert it to MPEG4, you cannot use a horizontal resolution higher than 720. Unless you zoom it, and it is not recommended.
So why is that? Why can't you use a higher resolution?

[SeeMoreDigital]

Thank you, SeeMoreDigital. Very nice site.
So if I understood, when using Anamorphic Widescreen, the actual resolution is 853*480 (in NTSC), and it is stored on the DVD as 720*480.
Then if I convert it to MPEG4 (AVI), I can use the resolution 848*480 for example. But I know that if you convert it to MPEG4, you cannot use a horizontal resolution higher than 720. Unless you zoom it, and it is not recommended.
So why is that? Why can't you use a higher resolution?The actual (resolution) pixel frame size on all commercial NTSC DVD's is 720 pixels wide by 480 pixel high. Without the aid of aspect ratio signalling embedded within the MPEG-2 video stream the pixels would be square and the displayed image would look squashed up upon playback. However, with the aid aspect ratio signalling the pixels become rectangular in shape and the displayed image looks normal upon playback.

As with an MPEG-2 video stream it's possible to add aspect ratio signalling to an MPEG-4 video stream. Meaning 720x480 MPEG-4 encodes, can look normal too upon playback.


Cheers

[ArielHatotah]

Nice to know it.
1. How can I add aspect ratio signalling to a MPEG-4 video stream?
2. Anyway, if I don't want to add aspect ratio signalling to my MPEG-4 video stream, and I want my MPEG-4 video stream to be 1:1 aspect ratio, then can I encode it with a horizontal resolution higher than 720, such as 800*432? Because the original resolution was 853*480, and has been squashed during encoding... Correct me if I'm wrong. Otherwise I don't understand the meaning of "anamorphic". (Anamorphic Widescreen is made to increase the resolution thus the quality)...

[SeeMoreDigital]

If you want to generate MPEG-4 encodes from MPEG-2 "wide-screen" DVD sources at "1:1", then your MPEG-4 encodes will have to be 720x480 (complete with 32:27 DAR signalling). Exactly like your "wide-screen" MPEG-2 DVD source.

To confirm.... The actual (resolution) pixel frame size on all commercial NTSC wide-screen DVD's is 720 pixels wide by 480 pixel high, complete with (32:27) DAR signalling embedded within the MPEG-2 video stream. The software player detects the DAR signalling and displays the video at (square pixel) resolution of 853x480 pixels. Even though the NTSC DVD source contains a fixed quantity of 720x480 pixels.

What is... an Anamorphic image? (http://www.seemoredigital.net/03_Video_Only_Info/What_is_an_anamorphic_encode.html)

[handtruck]

If you want to generate MPEG-4 encodes from MPEG-2 "wide-screen" DVD sources at "1:1", then your MPEG-4 encodes will have to be 720x480 (complete with 32:27 DAR signalling). Exactly like your "wide-screen" MPEG-2 DVD source.

To confirm.... The actual (resolution) pixel frame size on all commercial NTSC wide-screen DVD's is 720 pixels wide by 480 pixel high, complete with (32:27) DAR signalling embedded within the MPEG-2 video stream. The software player detects the DAR signalling and displays the video at (square pixel) resolution of 853x480 pixels. Even though the NTSC DVD source contains a fixed quantity of 720x480 pixels.


SeeMoreDigital,

So if I have a hardware player that doesn't recognize DAR signaling (and I do, a Cowon A2), what resolution would preserve the most of the "original" video information if I had to do 1:1 PAR and no DAR change. Would it be 853 x 480 or 720 x 405 (assuming 16x9 video)?

[SeeMoreDigital]

SeeMoreDigital,

So if I have a hardware player that doesn't recognize DAR signaling (and I do, a Cowon A2), what resolution would preserve the most of the "original" video information if I had to do 1:1 PAR and no DAR change. Would it be 853 x 480 or 720 x 405 (assuming 16x9 video)?If you're unlucky enough to own a hardware player that doesn't recognise aspect ratio signalling, then it's very likely it wont be able to recognise "width" resolutions greater than 720 pixels as-well.

Therfore you will have to re-size :(

[handtruck]

If you're unlucky enough to own a hardware player that doesn't recognise aspect ratio signalling, then it's very likely it wont be able to recognise "width" resolutions greater than 720 pixels as-well.

Therfore you will have to re-size :(

Actually, it goes up to 800, believe it or not.
But irregardless, for my personal collection, say I HAD to save a 1:1 PAR with no stated DAR from a 720x480 DVD. Would 853x480 be the way to go to preserve the original video?

[SeeMoreDigital]

Would 853x480 be the way to go to preserve the original video?It would be the perfect way to go....

Unfortunately though, quality encoding applications conform to "mod-16" sized resolution chunks. So the nearest you'll get to 853 pixels will be 848 or 864.


Cheers

[handtruck]

It would be the perfect way to go....

Unfortunately though, quality encoding applications conform to "mod-16" sized resolution chunks. So the nearest you'll get to 853 pixels will be 848 or 864.


Thanks for all your help!

One last question for you (I promise): Is there any quality loss at the same bitrate for 848x480 vs 720x480 (with the flag)? I'm a little confused as to how these are the "same" considering the big difference in number of pixels. I'm assuming it has to do with pixel SIZE, but I'm not sure.

[SeeMoreDigital]

Thanks for all your help!

One last question for you (I promise): Is there any quality loss at the same bitrate for 848x480 vs 720x480....Yes.... Generally, the more pixels you encode (61,440 or around 18% more in this case), the more bit-rate you'll require!

I'm a little confused as to how these are the "same" considering the big difference in number of pixels. I'm assuming it has to do with pixel SIZE, but I'm not sure.No, it's not the pixels size... it's the pixels shape. The pixels on a DVD have been designed to be viewed as rectangles, not squares ;)


Cheers

[handtruck]

Yes.... Generally, the more pixels you encode (61,440 or around 18% more in this case), the more bit-rate you'll require!

No, it's not the pixels size... it's the pixels shape. The pixels on a DVD have been designed to be viewed as rectangles, not squares ;)


Cheers

Sorry for the broken promise ;)

This is where I am still a bit confused, so bear with me. Even though the number of pixels are different, since the DAR should still be the same (848x480), shouldn't the bitrate required be the same?
I know the number of pixels are significantly different, but the end result (when viewing) is exactly the same, so if the 720x480 requires less bitrate as you say, there has to be some "loss" there.

Say I just do a generic xvid encode from a DVD at 720x480. I then use MPEG4Modifier to change the PAR to NTSC16x9. It will then display the video at 853x480, right? Does it do this by changing the shape, size, or both of the individual pixels before displaying. For it to go from 720 to 853, doesn't the size HAVE to change?

[ArielHatotah]

First of all, thank you SeeMoreDigital for all your help.
And I'm sorry to bother you, but I still didn't get it.
I know that
The actual (resolution) pixel frame size on all commercial NTSC wide-screen DVD's is 720 pixels wide by 480 pixel high
But before encoding it to MPEG-2, the original resolution was 853x480, and when encoded it has been changed "anamorphically" to 720x480. Then the player changes it BACK to its original resolution, 853x480.
So if I'm correct, when I want to encode it to MPEG-4, I can encode it with that original resolution (853x480 or something else to be mod-16).
Probably I'm wrong, but I didn't understand what is wrong...

And I also want to say that the group "WAF" encodes their videos with a high resolution of 800x336. Are they right? Or maybe their videos are mistakenly zoomed in...?

By the way, how do I make DAR signalling in MPEG-4?

[SeeMoreDigital]

Say I just do a generic xvid encode from a DVD at 720x480. I then use MPEG4Modifier to change the PAR to NTSC16x9. It will then display the video at 853x480, right? Does it do this by changing the shape, size, or both of the individual pixels before displaying. For it to go from 720 to 853, doesn't the size HAVE to change?With regard to software players. Once the shape of the "rectangular" or "non-square" pixels are detected by the player they're converted to a "square" pixel representation of the image.

[ArielHatotah]

After reading a little more, and thinking a lot about it, I think I understood. Now help me to know if I'm right.
When encoding to DVD (say NTSC), the designed resolution is 720x480, although the pixels are rectangular. So if I change it to 853x480, then I lose some quality, because I increased the resolution and stayed with the same bit rate (allocated by the DVD encoders).
Or maybe the designed resolution from the beginning is 853x480, and when encoding they use enough bit rate to fit this resolution. Therefore, when I encode to say MPEG-4, I can use a high resolution like 848x480 or so.

Which theory is true?

Thank you in advance.

[handtruck]

After reading a little more, and thinking a lot about it, I think I understood. Now help me to know if I'm right.
When encoding to DVD (say NTSC), the designed resolution is 720x480, although the pixels are rectangular. So if I change it to 853x480, then I lose some quality, because I increased the resolution and stayed with the same bit rate (allocated by the DVD encoders).
Or maybe the designed resolution from the beginning is 853x480, and when encoding they use enough bit rate to fit this resolution. Therefore, when I encode to say MPEG-4, I can use a high resolution like 848x480 or so.

Which theory is true?

Thank you in advance.

I've been thinking and researching a lot about this as well. When you look at a 720x480 video on the computer without any flags (such as in dgindex), you are viewing it with a PAR of 1:1. When you PLAY the video on your computer or TV, it CHANGES the PAR to 16:9 on playback. To do this, it is taking the square pixels and making them a rectangle by MAKING THEM LARGER (extending the width)- it HAS to, otherwise the height of the video would change. That's where the dilemma is - how does 720 become 853 without a change it quality?
What does this mean for MPEG-4 encoding? I'm not sure.. I'd love to continue this discussion some more. My goal is to figure out the resolution of an anamorphic mpeg-4 file (from a DVD) while maintaining a 1:1 PAR (Not enough and keeping as much info from the original video as possible. I don't think 848x360 (for 2.35:1) is the answer. I also don't think 720x306 is the answer either. Is there a happy medium?

Let's investigate - If you have a 2.35:1 video on an NTSC DVD, the actual video size (subtracting the black bars) is usually something like 720x368. So how do we make that 2.35:1? Let's try some math:
720x368 = 264960 pixels.
Height x (2.35 x height) = 264960
height = 335.8; therefore
width = 789.1
using mod16 is approximately 784x336

In my opinion, THIS should be the dimensions of an MPEG-4 2.35 encode. It preserves the FULL resolution of the original file.

I think this has to be done simply because of the lack of ARS support out there.

Any other thoughts?

[SeeMoreDigital]

I've been thinking and researching a lot about this as well. When you look at a 720x480 video on the computer without any flags (such as in dgindex), you are viewing it with a PAR of 1:1. When you PLAY the video on your computer or TV, it CHANGES the PAR to 16:9 on playback. To do this, it is taking the square pixels and making them a rectangle by MAKING THEM LARGER (extending the width)- it HAS to, otherwise the height of the video would change. That's where the dilemma is - how does 720 become 853 without a change it quality?The displayed image doesn't actually become 853 pixels wide. It's just the software player that does this!

As I mentioned before, the software player makes a "square" pixelled representation of the original "non-square" (rectangular) pixelled image.... It's just something that software players have to do!


What does this mean for MPEG-4 encoding? I'm not sure.. I'd love to continue this discussion some more. My goal is to figure out the resolution of an anamorphic mpeg-4 file (from a DVD) while maintaining a 1:1 PAR (Not enough and keeping as much info from the original video as possible. I don't think 848x360 (for 2.35:1) is the answer. I also don't think 720x306 is the answer either. Is there a happy medium?

Let's investigate - If you have a 2.35:1 video on an NTSC DVD, the actual video size (subtracting the black bars) is usually something like 720x368. So how do we make that 2.35:1? Let's try some math:
720x368 = 264960 pixels.
Height x (2.35 x height) = 264960
height = 335.8; therefore
width = 789.1
using mod16 is approximately 784x336

In my opinion, THIS should be the dimensions of an MPEG-4 2.35 encode. It preserves the FULL resolution of the original file.Your nearly there... but you've neglected to remember that MPEG-4 encoded pixels can be "non-square".

I think this has to be done simply because of the lack of ARS support out there.Actually, most software players and MPEG-4 direct-show decoders offer ARS support now.... As do most good quality hardware stand-alone players


Cheers

[handtruck]

Your nearly there... but you've neglected to remember that MPEG-4 encoded pixels can be "non-square".

So if you accidentally encode an MPEG-4 file at 720x480 at the default 1:1 PAR, and then use MPEG4Modifier to change the PAR, are you actually changing the pixels to "non-square" or are you putting a flag in the stream to tell the playback device to display the video at the new PAR.
Is this different than setting the PAR at encode time (such as the xvid_encraw -par switch)?

[SeeMoreDigital]

So if you accidentally encode an MPEG-4 file at 720x480 at the default 1:1 PAR, and then use MPEG4Modifier to change the PAR, are you actually changing the pixels to "non-square" or are you putting a flag in the stream to tell the playback device to display the video at the new PAR.Yep... you're assigning the "square" pixel a "non-square" pixel shape. Essentially replicating how the MPEG-2 DVD source does it.

Is this different than setting the PAR at encode time (such as the xvid_encraw -par switch)?It's the same...

[SealTooGreat]

So if you accidentally encode an MPEG-4 file at 720x480 at the default 1:1 PAR, and then use MPEG4Modifier to change the PAR, are you actually changing the pixels to "non-square" or are you putting a flag in the stream to tell the playback device to display the video at the new PAR.

MPEG4Modifier just assigns the flag in the stream which tells the player to display the video at the new PAR. Actually, your source is still at resolution 720x480, only PAR flag has been changed. (another interpretation of what SMD said)

Changing pixel shape from square to non-square is upon playback.

[ArielHatotah]

I see that there is a disagreement. And it's important to know how this works. There is a significant difference.
So how is it in a DVD? Are the pixels square and being rectangular during playback? Or the opposite?
And how is it in MPEG4Modifier? Does the software just put a flag? Or change the shape of the pixels? And is it the same like when encoding?

By the way, is "handtruck" right? Because he seems to be right. I agree with the math he did. (Only the actual video size (subtracting the black bars) is 480 * 16 / 9 / 2.35 = 363, therefore 720x363, but it is not big deal).
And if what you said, handtruck, is true, you can see that you always result with width = √614400 = 783.8367
and hight= 783.8367 / AR, where AR is the actual Aspect Ratio (in your example it's 2.35).
And that's because:
The number of pixel in the DVD is 720 * (480 * 16 / 9 / AR) = 614400 / AR.
So (width / AR) x width = 614400 / AR.
Thus width = √614400 = 783.8367

And if it is true, then I don't understand why you can't encode in a higher horizontal resolution than 720. (Look at this (http://www.planetnana.co.il/arielhatotah/The.Matrix.Revolutions.2003.DVDRip.XviD.AC3%2DFH%2DPROPER.txt), at the comment on WAF release (their release is at a resolution of 800x336).

[Wilbert]

I see that there is a disagreement. And it's important to know how this works. There is a significant difference.
So how is it in a DVD? Are the pixels square and being rectangular during playback?
A pixel doesn't have a shape. It is square nor rectangular.

When your source is 720x480 (NTSC DVD) and IF the DAR (MPEG-2 doesn't store a PAR flag in its header, just an DAR flag) would be 1:1, it means your video is played back at 720x480, while it should be 853x480 for anamorphic video. This means that your played video is horizontally squeezed (this is what the posters mean with rectangular pixel shape here, but again pixels have no shape).


SealTooGreat said it correctly (except for his last sentence):
MPEG4Modifier just assigns the flag in the stream which tells the player to display the video at the new PAR. Actually, your source is still at resolution 720x480, only PAR flag has been changed. (another interpretation of what SMD said)

And if it is true, then I don't understand why you can't encode in a higher horizontal resolution than 720.
You can, but it at a higher bitrate to get the same quality (resulting in a larger filesize). If you don't care about filesize you can do this, but you can also let it be done by the resizer upon playback. If the PAR is not equal to 1, but 16/9*480/720 = 1.1852 in your case, your video is just resized horizontally by 1.1852, getting 1.1852*720 = 853 upon playback. (Sorry for the ugly sentence :))

[ArielHatotah]

And if it is true, then I don't understand why you can't encode in a higher horizontal resolution than 720.
You can, but it at a higher bitrate to get the same quality (resulting in a larger filesize). If you don't care about filesize you can do this, but you can also let it be done by the resizer upon playback. If the PAR is not equal to 1, but 16/9*480/720 = 1.1852 in your case, your video is just resized horizontally by 1.1852, getting 1.1852*720 = 853 upon playback. (Sorry for the ugly sentence :))

O.K, so what is the original resolution encoded on the DVD?
Let's phrase it like that: If I want to encode with the same bitrate as the original DVD has, without losing nor gaining any quality, what resolution should I use for my video?
Is it 720x405 (assuming the Aspect Ratio is 16:9)? Is it 853x480? Or 784x441 according to handtruck's computing...?

And again, why are WAF's releases considered zoomed in? (Look here (http://www.planetnana.co.il/arielhatotah/The.Matrix.Revolutions.2003.DVDRip.XviD.AC3%2DFH%2DPROPER.txt) at the comment on WAF's release (their release is at a resolution of 800x336)).

[foxyshadis]

If you want to encode at the same bitrate, keep the DVD! Transcoding just for the sake of transcoding is pointless.

To the main point, though, you just have to decide, and perhaps try a few combinations. You've been given all the information in this thread to do it; you can encode 1:1 or anamorphic, you can encode by expanding the horizontal, shrinking the vertical, or a combination of both. The 'best' option depends on how you play it back (for instance, with Haali's renderer or VMR9 you can use bicubic, which looks much sharper than that generic bilinear otherwise used), how much space you allot (the lower the bitrate, the less pixels you want, to maximize quality of coarse detail; this ties into bpf, but the sweet spots are different for everything), and how you resize (lanczos will add a lot more noise to the video compared to bilinear at the same bitrate, because of that sharpness). What's most important is figuring out how you want the end result to look, because there are so many options. We don't have your eyes, your TV, or your preferences.

However, you cannot ever shrink the video without losing some quality, unless the movie was upscaled (or out of focus) in the first place. Similarly, upscaling can't add detail, at best they can make existing detail easier to see. Which is why if you aren't starved for bitrate (1cd or even 2 for many movies), and have the ability to play it back with bicubic or lanczos resizing, anamorphic is a really smart way to go.

I've never heard of WAF or seen their releases, without further clarification it's impossible to say. Maybe you should ask this stereozulu?

[SeeMoreDigital]

O.K, so what is the original resolution encoded on the DVD?.....As I previously mentioned (post 6). The actual resolution (pixel frame size) on all commercial NTSC DVD's is 720 pixels wide by 480 pixels high - The actual resolution on all commercial PAL DVD's is 720 pixels wide by 576 pixels high!

Upon playback, the square pixels are re-shaped to non-square. So the displayed image looks the correct proportions.

[Wilbert]

Let's phrase it like that: If I want to encode with the same bitrate as the original DVD has, without losing nor gaining any quality, what resolution should I use for my video?
Is it 720x405 (assuming the Aspect Ratio is 16:9)? Is it 853x480? Or 784x441 according to handtruck's computing...?
To add some comments. With lossy re-encoding you will always loose some detail. Having said that (assuming you re-encode at the same bitrate as the original) ...

Is it 720x405 (assuming the Aspect Ratio is 16:9)?
Since this has less pixels than the original, you won't loose that much quality. So, this is a good choice.

Is it 853x480?
Since this has much more pixels than the original, you will loose much quality here.

Or 784x441 according to handtruck's computing...?[
Assuming these numbers are correct. This has about the same amount of pixels as the original. But the difference is that many pixels in the original are black (the black borders) and in your encoding you don't have black borders, which implies that you will loose quality.

So, in your case, i would choose your first option.

[SealTooGreat]

...but again pixels have no shape..
SealTooGreat said it correctly (except for his last sentence):

Take look at Pixel Aspect Ratio - Wiki (http://en.wikipedia.org/wiki/Pixel_aspect_ratio)

http://upload.wikimedia.org/wikipedia/commons/7/73/PAR-1to1.jpg

http://upload.wikimedia.org/wikipedia/commons/a/ac/PAR-2to1.jpg

[Wilbert]

The pictures have nice colors, but are misleading. They don't correspond with what is stated in the text:

The term pixel aspect ratio, sometimes shortened to aspect ratio, is also used in the context of computer graphics to describe the distribution of pixel in a digitized image.

I encourage you to read the document: A Pixel is Not a Little Square (ftp://ftp.alvyray.com/Acrobat/6_Pixel.pdf).

You might read the analog capture guide (http://www.doom9.org/index.html?/capture/par.html) too. But since we wrote this, i guess you won't accept it as a higher authority.

[SealTooGreat]

I have always found this wiki's graphical Pixel shape interpretation a bit weird, cause upon higher resizing there's interpolation which add new pixels, but not stretch original ones, so the right word would be distribution of pixel. I haven't read documentations you gave, yet. It seems interesting ones, especially "A Pixel is Not a Little Square", though I know about "analogue capture guide" before, but not in the sense of PAR explanation. thnx :)

edit: PAR tells you the shape of the pixel distribution or pixel grid in the video (from analog capture guide (http://www.doom9.org/index.html?/capture/par.html)).
Indeed, pixel grid really gives the graphical sense of pixel distribution. :)

[handtruck]

But the difference is that many pixels in the original are black (the black borders) and in your encoding you don't have black borders, which implies that you will loose quality.


My calculation was after removing the black borders.

I will maintain that in the absence of players that acknowledge the Aspect Ratio flag in an MPEG-4 stream (which is an unfortunate truth), the alternative which will yield the same resolution as the original DVD is to:
-remove any black bars
-follow this formula:
(NewHeight) * (NewHeight)(DAR) = (OldHeight)(OldWidth)

Please note OldHeight and OldWidth are what results after removing the black bars, and (NewHeight)(DAR) = NewWidth.

This isn't as hard as it looks (see my previous posts for this in action).

IMO, this is the best way to do a DVD backup regardless of codec. You ensure full DVD resolution, and almost universal playback.