still confused about PAR

[Arachnotron]

Any PAL DVD which contains a movie with an aspect ratio of 4:3 (1.33:1), contains an Mpeg2 stream which occupies all 720x576 pixels - all image, none matte.

If you look on this (http://gregl.net/videophile/anamorphic.htm) page, you will find that there are some odd-ball movie formats which result in anamorphic DVD's on which not all pixels contain image and some do contain matte.

(by the way, I don't like their interpretation of the term 'overscan')

There actually exists an ITU standard about all the movie sizes there are and how to position the frame. (forgot which, maybe Trev or Wilbert still have the reference?)

[SeeMoreDigital]

Originally posted by Wilbert
Give me one reference which uses anamorphic in this way. You've probably already got some PAL DVD's which are '4:3' (or 1.33:1) aka Std Frame, aka Full Frame on one side and 'widescreen' on the other!


Cheers

[Wilbert]

You've probably already got some PAL DVD's which are '4:3' (or 1.33:1) aka Std Frame, aka Full Frame on one side and 'widescreen' on the other!
True (assuming you mean anamorphic widescreen), and gras is green. I give up.

[Arachnotron]

In case anyone is interested: I found the ITU references back again.

RECOMMENDATION ITU-R BR.1374-1*
Scanned area dimensions from 16 mm and 35 mm cinematographic film used in television

and maybe

ITU-R BT.1379-1
Safe areas of wide-screen 16:9 and standard 4:3 aspect ratio productions to achieve a common format during a transition period to wide-screen 16:9 broadcasting

are interesting in the light of this discussion.

You can download 3 for free each year after registration.

http://www.itu.int/home/index.html

[Stux]

Oh, Boy, lots of confusion.

Lets be clear, Anamorphic is a term which comes from the world of cinema. An anamorphic lens is one which does not project true, and in fact elongates the image horizontally (or squishes vertically, depending on your perspective) to produce a wide image from a squarish (approximately 4:3) image.

Approximately 4:3 -> 16:9, or as they like to say in the cinema world, 1.33:1 -> 1.85:1

Yes, 16:9 is not exactly 1.85:1, whatever, 16:9 is not exactly 16:9 either... which is one reason PARs are used.

The other way to produce a 1.85:1 image from a 1.33:1 image is to insert mattes into the projected image to crop off the top/bottom... but then you are wasting film

Then we move a long, 1.85:1 evolved to 2:1 and then 2.35:1

2.35:1 is normally made by matting an anamorphic 1.33:1 image.

...

<insert a bunch of stuff about the world of video and tv and D1 and widescreen and mpeg-2, and telecines which I can't be bothered typing>

...

Which leads us to DVD, where there are two Display Aspect Ratios defined, 4:3 and 16:9.

When a video is encoded with the 16:9 flag set it is said to be Anamorphic.

There is no 4:3 Anamorphic.

[Kika]

OK, some thoughts...

Does Pixels have a shape?
Not really, but to keep the math (and the thinking about PAR/AR) simple, you can treat them as shaped. That's what PAR means.

What does anamorphic mean?
In it's simple form, it means a squeezed image. But like Stux wrote: it came from the world of Cinema and should used only for 16:9 anamorphic Video.

What resolution really is 4:3?
Um, that's a bit hard. Regarding to Signal-Timings, PAL 4:3 is 702x576, NTSC is 711x486.
Regarding to MPEG-Videos and mod16 rules, PAL is 704x576, NTSC is 704x480.
In the World of square Pixels, things are different. PAL 4:3 is 768x576, NTSC 640x480 (but the TRUE NTSC-Resolution is 648x486!)
But damn, what is 720x480/576 for?
It is exactly the same like 704x480/576 but using a amount of 16 extra Pixels. This extra Area isn't clearly defined. It can contain black bars, it can contain image, but it should never be visible on any TV. So any kind of Video-Resolution using the full 720 pixels for an image never has the exact AR (DAR) auf 4:3!

What is this damn PAR-Thing? MPEG uses non square Pixels, the DAR can (and should) be used as a factor to stetch the horizontal Image out to it's square pixel resolution to get the concrete AR.
As an Example: In PAL 4:3, the PAR is 1.092:1 (1.094:1).
So 704*1.092=~768
But 720*1.092=~786
4:3=1.33
768/576=1.33 (perfect 4:3)
But 786/576=1.36! (that's NOT 4:3)

What do you think?

[SeeMoreDigital]

OK

Lets talk PAL.

Encode a 16:9 (1.77:1) 'image' using all 720x576 pixels. The image is 'squashed/squeezed' horizontally to occupy all the pixels. So it therefore stored 'anamorphicly... yes?

Now encode a 4:3 (1.33:1) 'image' using all 720x576 pixels. The image is also 'squashed/squeezed' horizontally to occupy all the pixels....

So bearing in mind that the encoded image has been 'distorted' to fit the pixels... what would you guys call this?


Cheers

[Kika]

@SeeMoreDigital

That's simple. 720x576 16:9 anamorphic contains an anamorphic image! The image itself is squeezed, the pixels are squeezed too.

720x576 4:3 contains an 4:3 non anamorphic image, but pixel which are non square.

Both of them are using the same PAR of 1.092:1 with - if they are propper resized and encodet - an motion (active) area of 704x576.

To correct my own witings: the term anamorphic should be used for image resolutions only. That's much easier to handle.

[bb]

I think whether or not the horizontal number of pixels equals 704 or 720, the line length to which it gets expanded is the same. So I assume it doesn't matter if you resize a square pixel 640x480 video (let's assume it's progressive) to 704x576 or 720x576. During playback both videos should be resized to a proper 4:3 picture, which means the 704x576 video would be stretched a little more horizontally.

If this assumption is false, then the 640x480 video (PAR 1:1) should be cropped at the left and right sides before resizing to 704x756, but it should not be cropped before resizing to 720x576.

There's a way to verify this (I've done this for DV material): Use a 640x480 test video with PAR 1:1 with a large square in the middle of the picture - it's easier to measure squares than circles ;) Resize to 704x576 and 720x576, burn it to DVD and measure on your TV screen whether the squares have turned into non-square rectangles or not.

The same applies to NTSC, just replace 576 by 480.

bb

[Arachnotron]

What resolution really is 4:3?
Um, that's a bit hard. Regarding to Signal-Timings, PAL 4:3 is 702x576, NTSC is 711x486.
Regarding to MPEG-Videos and mod16 rules, PAL is 704x576, NTSC is 704x480.

Not exactly. VCD and DVD players use fixed sample rates when playing back these files. as such, only 702x576 results in a 52 µs 'wide' signal. choosing 704x576 has indeed to do with it being mod16, but it is not exactly 52 µs wide (it is 52.14) and as such does not results in a 4:3 frame. It is a bit wider. However, on a TV the difference is in the overscan area and you will never notice. Software players, especially when they are sloppy with the norms, are another thing. And I suppose modern TV's that can size right up to the outer edge of the picture and 'sense' letterboxing are another problem.

In the World of square Pixels, things are different. PAL 4:3 is 768x576, NTSC 640x480 (but the TRUE NTSC-Resolution is 648x486!)
But damn, what is 720x480/576 for?

My guess is this has to do with acommodating differences in the position of the center in the analogue signal. Once you choose 13.5 MHz sample rate AND have to accomodate both NTSC and PAL you need a bit of leeway at the sides.

[Arachnotron]

I think whether or not the horizontal number of pixels equals 704 or 720, the line length to which it gets expanded is the same.


This is incorrect!!!
A DVD player operates at a fixed sample rate of 13.5 MHz. Each line of pixels results in an analogue video scanline.
720 pixels/ 13.5 MHz = 53.3333 µs
704 pixels/ 13.5 MHz = 52.14 µs

One source of confusion is what MPEG2 encoders do when you feed them non-720 resolution files. Some will letterbox, some will resize.
Depending on the setting, a 704x576 source can result in:
stretched to 720x576 MPEG2(wrong AR)
letterboxed to 720x576 MPEG2(correct AR)
not resized, still 704x576 MPEG2 (also correct AR)

a single pixel on a DVD will always result in (1/13.5) µs of analogue signal

[Kika]

a single pixel on a DVD will always result in (1/13.5) µs of analogue signal Yes, no matter, what the resolution of the Video is.

[Arachnotron]

Yes, no matter, what the resolution of the Video is.
correct. Though there is not much to choose from there, only 704 and 720 are valid horizontal resolutions.

For SVCD and VCD the story is pretty much the same.
SVCD is 9 MHz, so 480/9 = 53.333 µs
VCD is 6.75 MHz, so 352/6.75 = 52.14 µs

If you go back to the analogue signals this stuff is not so difficult to get right. Only, nowadays many people watch everything on a PC, and there is no overscan area anymore to hide the differences in.

[edit]of course, 352x576 DVD is not played back at 13.5 MHz but at 6.75

[Kika]

only 704 and 720 are valid horizontal resolutions.

Um, don't forget Half-D1 or Quarter-D1 which are valid resolutions too.

[SeeMoreDigital]

Originally posted by Kika
That's simple. 720x576 16:9 anamorphic contains an anamorphic image! The image itself is squeezed, the pixels are squeezed too.

720x576 4:3 contains an 4:3 non anamorphic image, but pixel which are non square. Agreed

Originally posted by Kika
To correct my own witings: the term anamorphic should be used for image resolutions only. That's much easier to handle. But image resolutions stored how? As lines, dots per square inch or as we do now using pixels, that as we all know may be distorted too!

Whether it's the image that's squeezed, or the pixel that's squeezed. One way or another the it's squeezed. And it's shape has to be changed/distorted to make it look the correct shape!


Cheers

[trevlac]

Originally posted by SeeMoreDigital
Whether it's the image that's squeezed, or the pixel that's squeezed. One way or another the it's squeezed. And it's shape has to be changed/distorted to make it look the correct shape!


You are getting it wrong here. :) Pixels have no shape. They can not be 'squeezed'. They are points. PAR is bad. Math is bad ;)

Picture this:

(A) ____________
(B) . . . . . . . . . .
(C) ...................

A = Analog signal line
B = Line sampled at 702 'pixels'
C = Line sampled at 768 'pixels'

768x576 and 702x576 are both 4:3 framed images. C just has more sample points. There is no stretching/squeezing of images/pixels.

BTW: C (768x576) is called square because the space between two horizontal sample points and the space between two vertical sample points is the same.

Another picture (I saw this somewhere):

(1) Square sample grid
. . . . . . .
. . . . . . .
. . . . . . .

(2) Non square sample grid
.............
.............
.............

They cover the same picture.


@bb,

To really go from a 640x480 square you need to resize to 711x486 and crop/pad to 704x480. Strangely enough, if you resize to 640x480 -> 702x480 and pad to 704x480, you get about the same result.

@Arachno,

OK I had to use PAR. 486 / (711*3/4) == 486 / 533.25 == .911 PAR
640 / .911 ~ 702.

Here is a reference with all of those numbers: http://www.quantel.com/domisphere/infopool.nsf/html/dfbAspectRatio?OpenDocument

[Arachnotron]

@trevlac
OK I had to use PAR. 486 / (711*3/4) == 486 / 533.25 == .911 PAR
640 / .911 ~ 702.

Er... what are you trying to calculate here?
[edit] nerver mind :)

[Kika]

@trevlac

OK, ok, But do you think that's easier for a Newbie?

I know, Pixel does not really have a shape, but why no use this imagination for doing some maths? I think it is much easier to think about shaped Pixels than using your correct but a little bit complex mathematics.

[Arachnotron]

I know, Pixel does not really have a shape, but why no use this imagination for doing some maths? I think it is much easier to think about shaped Pixels than using your correct but a little bit complex mathematics.

It's not that complex if you do it step by step and throw in some explanations. I'm sure Trevlac will be happy to oblige. ;)

[Kika]

I'm sure Trevlac will be happy to oblige.

Oh, OK. trevlac, i think it's your turn now... ;)


@SeeMoreDigital
I don't want to make the confusion complete. But is is also an imagination to think of 720x576 4:3 as a squeezed Image. It will be shown as a squeezed image on your monitor if you are using a simple application which isn't able to do the corrections.
If PC-Monitors where working like TVs do, we all might have much less problems with this PAR/DAR/AR (whatsoever) stuff. :rolleyes:

[SeeMoreDigital]

Originally posted by Kika
I don't want to make the confusion complete. But is is also an imagination to think of 720x576 4:3 as a squeezed Image. It will be shown as a squeezed image on your monitor if you are using a simple application which isn't able to do the corrections.
If PC-Monitors where working like TVs do, we all might have much less problems with this PAR/DAR/AR (whatsoever) stuff. :rolleyes: Well I think you've answered one of the questions...

If you need something to 'correct' the image, whether it be software, mirrors, lenses or even David Blaine. That means the image at some time has been distorted!


Cheers

[trevlac]

Originally posted by SeeMoreDigital
Well I think you've answered one of the questions...

If you need something to 'correct' the image, whether it be software, mirrors, lenses or even David Blaine. That means the image at some time has been distorted!


Again, I respectfully disagree. :)

The image is stored. Then it is displayed.

A device displays it based upon an known standard. A hardware DVD player knows the digital sample rate of the file, and the NTSC/PAL analog signal standard. Although there are more layers on a PC, a Software player knows the digital sample rate of the file, and knows how the graphics card will create a VGA signal.

So ... a digital image is not distorted due to pixels. 1 Pixel is not a little cut out piece of a picture.

There is no correcting involved. Either the player knows the standards or it does not. Either it can properly 'recreate' the stored image, or it can not.

Here is a paper that goes into this a bit....

http://www.cs.princeton.edu/courses/archive/fall00/cs426/papers/smith95b.pdf

Here is something that someone wrote to refute the paper.
http://www.csbruce.com/~csbruce/cynic/pixel.html

Even though I believe the 1st, there is some merrit to the 2nd. Sometimes when you make pixels, there is a thing called the appature effect. The measure is really an average over some period of time. However, the thing missing from the 2nd article is that this measuring period of time is not the complete period of time until the next pixel measurement. So the apature does not fill up the space between pixels.

[SeeMoreDigital]

Do you agree any of the following: -

01: That the shape of the 'stored' image is not the shape of the 'displayed' image?

02: That in order to view the stored image correctly (weather it be 4:3 or 16:9 image), some form of manipulation takes place?

If the answer is yes to both questions then by definition the image is anamorphic.

Even images viewed through a 'fish eye' lens (like the spy hole in your front door) can be classed as anamorphic images.

The Oxford English dictionary, says: -
anamorphosis (noun): a distorted image which appears normal when viewed from a particular point, or with a suitable mirror or lens. The process by which such images are produced.
-DERIVATIVES: anamorphic (adjective)
-ORIGIN: from the Greek anamorphosis 'transformation'


The Odhams English dictionary, says: -
ANAMORPHOSIS (noun): distorted representation of an object which appears in correct proportion when looked at from a certain angle or when reflected from an convex or concave mirror or lens.


The Merriam-Webster Online Dictionary, says: -
Main Entry: ana·mor·phic
Pronunciation: "a-n&-'mor-fik
Function: adjective
Etymology: New Latin anamorphosis distorted optical image
: producing, relating to, or marked by intentional distortion (as by unequal magnification along perpendicular axes) of an image <an anamorphic lens>


The Encyclopædia Britannica, says: -
In the visual arts, an ingenious perspective technique used to give a distorted image of the subject represented in a picture when seen from the usual viewpoint, but so executed that if viewed from a particular angle, or reflected in a curved mirror, the distortion disappears and the image in the picture appears normal.



Cheers

EDIT: changed the word actual to displayed

[Wilbert]

In the digital video world the term anamorphic is used as in "16:9 anamorphic widescreen", suggesting otherwise because only the dictionary says so is simply misleading to everyone involved.

Btw, the image has no shape when it is stored, only when it is displayed. Of course, it constains information how it should be displayed. Sometimes this is correct, sometimes not. Sometimes the player handles it correct, sometimes not. By the "common" definition of anamorphic (in the video world), this has nothing to do with being anamorphic.

Even if you want to use it in that way you come into conflict. Suppose you got a 16:9 widescreen video (which is called anamorphic using the common definition). Suppose that it is played on a 16:9 screen (in a correct way, as it should). By your definition it would not be anamorphic because it is not distorted.

[trevlac]

Originally posted by Kika
@trevlac

OK, ok, But do you think that's easier for a Newbie?

I know, Pixel does not really have a shape, but why no use this imagination for doing some maths? I think it is much easier to think about shaped Pixels than using your correct but a little bit complex mathematics.

Actually ... my math was with PAR. Without PAR, you can use Jukka's table for the work.

We will use bb's example: 640x480 PC -> 720x480 NTSC DVD

1. Make the source height match the target "actual active picture" height. Go to Jukka's table for the values. DVD = 486(H) 640x480 -> 648x486. You could do this by padding or resizing. Regardless, you need to keep the perspective as 4:3.

2. Go to Jukka's table and get the sample rates. 648x486 (Computer format) is 12.3064 and 720x480 (NTSC TV) is 13.5.

3. Find the 'line length' covered by the source width 648 / 12.3064 = 52.655.

4. Resample (resize) the picture line length to your target standard 52.655 * 13.5MHz = 711 (AKA resize 648x486 to 711x486)

5. Crop/pad to your target 711x486 -> 720x480

-----------------------------
The 1st step is messy for NTSC. For PAL, this is not an issue.

Arachno could object that this method only covers the standards listed in Jukka's page. Whatever he says, he knows more than me, and usually can state it more clearly.

[trevlac]

@SeeMoreDigital

I agree with what Wilbert said.

In the digital video world the term anarmorphic is used as in "16:9 anamorphic widescreen", suggesting otherwise because only the dictionary says so is simply misleading to everyone involved.

Btw, the image has no shape when it is stored, only when it is displayed. Of course, it constains information how it should be displayed.

Therefore, I don't agree with your #1. A stored image has no 'stored' shape. It has info about what shape to use when you display it. Just like it has no color.

Maybe you could restate your point:
- Pixels are anamorphic?
- There is a thing such as 4:3 anamorphic?

[edit]
Wilbert beat me to it! I was just going to ask how SMD describes the operation of a Widescreen TV.

[SeeMoreDigital]

The term '16:9 anamorphic widescreen' is misleading in itself.

The term 'anamorphic wide-screen' should be used (as should 'anamorphic std-screen' in my opinion :D).

Okay, it's clear we are never going to agree about this but one thing is clear, many of the definitions and abbreviations used are confusing.

Indeed some of the definitions used with regard to discussing 'film' do not carry over well when discussing 'digital' topics.

Now can anybody tell me which ones of the following spellings they prefer. ANALOG or ANALOGUE?


Cheers

[stephanV]

WATCH OUT! noob at work! :devil:

(I'm confused so i hope someone will correct me if I'm wrong :) )

i think the cause of the confusion is the difference between the resolution (=number of pixels?) of a movie (e.g. 720x576, 720x480) and the (dare i say it? :D ) aspect ratio of a movie (e.g. 4:3, 16:9). these two things are not the same as i see it now.

if i understand it correctly, how a movie will be shown on a screen depends on the shape of the pixel-grid (i guess this is what PAR means) and NOT AT ALL on the resolution as such. in theory, a movie with a resolution of 280x576 still can have a aspect ratio of 4:3, no? it will only depend on the shape of the pixel-grid how the video will be shown. and as the pixel-grid of TV and a computer-screen have different shapes and videos/DVDs are originally made for TV a computer screen will display the video in the wrong AR, when no action is taken to correct this.

now 'anamorphic wide screen' must mean then that when a movie with a resolution of 720x576 and a AR of 16:9 will be displayed on a normal 4:3 TV the image will appear to be squeezed (therefore we need letter boxing.)

does any of this make sense to anyone?

PS. i prefer analog ;)

[Arachnotron]

@stephanV

Agreed (apart from the nitpicking comment a TV has continuous lines, no pixels at all :) )

And I prefer analogue; I'm a snob :D

[edit] so for TV, a row of pixels first has to be translated to a scanline by the player.

[stephanV]

Originally posted by Arachnotron
[edit] so for TV, a row of pixels first has to be translated to a scanline by the player.

hmmm, but thats still confusing for me :)

i think when i said this:

resolution (=number of pixels?)

it would have been better to say 'samples' instead of 'pixels', right?

so a row of samples gets translated into a scanline by a TV and into a row of pixels by computer screen! (or am i mixing things up now? :rolleyes: )

[Arachnotron]

so a row of samples gets translated into a scanline by a TV and into a row of pixels by computer screen! (or am i mixing things up now? )

We ( or I am :) ) digressing a bit, but:
A DVD player reads the pixels/samples one by one and row by row at a fixed rate, and converts them into a voltage that fluctuates according to the information in the pixels. This fluctuating voltage signal, also known as composite/svideo/RGB :) in turn controls the electron beam in the Tv set and causes it to draw the picture line by line.

(some extra synchronization lines and pulses are added too, but basically that is it)

when you put a VCD in the player, it simply reads the samples more slowly so a single line finishes in about the same time as for a full size DVD.

A software DVD player emulates both the DVD player AND the TV electronics.
It calculates what the picture would look like on a TV screen, taking into account things like AR and letterboxing, deinterlaces on the fly, resizes the frame to whatever resolution your PC is running at and displays it one frame at a time.

[stephanV]

Originally posted by Arachnotron
We ( or I am :) ) digressing a bit
that might be, but your last post is one of the few posts in this thread i could actually understand.

many thx :)

[SeeMoreDigital]

Originally posted by stephanV
that might be, but your last post is one of the few posts in this thread i could actually understand.

many thx :) Cheeky... you'll get it... eventually!


Cheers

[Kika]

Guess we have found a set of terms and mathematics now that suits us all. Isn't it time for a summary or a guide?
I think this thread will kill newbies. ;)

BTW: I prefer Analog, because it's the same word in english and in german. :D

[Arachnotron]

I think part of the problem is where people come from. For me, TV is still primarily an analogue affair. I tend to think in voltages, scanlines and microseconds first. VCD/DVD are for me just places to store these analogue signals in digital form before converting them back to analogue signals for my TV set.

But if you come in from the DivX side of things, i.e. computer only, you never leave the digital domain. It can be very difficult to get your brain around things like overscan, aspect ratio and anamorphic since these terms are rooted in the limitations of analogue equipment.
Things like 704 and 720 DVD resulting in the same image on a TV set are easy to understand if you look at it from the analogue side but baffling for a computer graphics only person.