Aspect Ratio question [Archive]

kurkosdr

10th May 2011, 16:58

When viewed on a monitor with square pixels (which is practically all computer monitors and all LCD TV screens), a 16:9 PAL DVD must be 1024x576. This is because (576*16)/9 = 1024, aka 1024x576. For 4:3 PAL DVDs, it's (576*4)3 = 768, aka 768x576. Pixels that don't contain picture information (such as blanking) don't count.

Now, with 16:9 NTSC DVDs, we get a (480*16)/9= 853.3333. Which is not a whole number! But convention has it that we take the next integer number, which is 854, aka 854x480. For 4:3 NTSC DVDs, we get a (480*4)/3 = 640, aka 640x480. Does this mean that when we view 4:3 NTSC DVDs on our computers, we see less horizontal detail that the DVD has? Some more knowledgeable member enlighten us!!

Now, some movies have black bars on top and bottom hardcoded on the picture (as to simulate a 21:9 aspect ratio on a medium like DVD which supports only 4:3 and 16:9). These black parts are (technincally) part of the picture, and do count towards the size.

Ghitulescu

10th May 2011, 19:30

1024 is for getting back, 1050/1048 is for gettin in

:p

In other words: 1024 is the right answer coming back from DVD
1050 should be (I personally use 1024) the right answer for CGI designed for DVD.

Sharc

10th May 2011, 20:12

Hello,
I needed to know how's that PAL DVD video is calculated display size as 1047x576, instead 1024x576, when arithmetically 1024 is correct (576 * 16 / 9 = 1024).
The video has Display Aspect Ratio attribute set to 16:9.
It depends on how the studio has authored the PAL DVD.
- If they followed the ITU pixel shape, the 720/576 anamorphic DVD picture should be displayed at (512/351)x720/576=1050/576.
- If they followed the MPEG4 pixel shape, it should be displayed at (16/11)*720/576=1047/576.
- If they followed the 'generic' pixel shape (mostly this is the case), it should be (64/45)*720/576=1024/576.
Only the studio could tell how they actually did it. It's not published. The difference is however less than 3%, so who cares ....

Anakunda

30th July 2011, 13:29

It depends on how the studio has authored the PAL DVD.
- If they followed the ITU pixel shape, the 720/576 anamorphic DVD picture should be displayed at (512/351)x720/576=1050/576.
- If they followed the MPEG4 pixel shape, it should be displayed at (16/11)*720/576=1047/576.
- If they followed the 'generic' pixel shape (mostly this is the case), it should be (64/45)*720/576=1024/576.
Only the studio could tell how they actually did it. It's not published. The difference is however less than 3%, so who cares ....

Yet I have to ask which pixel shape gives a true 16:9 shape during playback? Does it soomething to do with monitor pixel shape? Didnot measure the video window on the screen as I don't have exact measurement equipment but I prefereably like to use the aspect ratio that gives true 16/9 shape on monitor. This is because I meet many videos in AVI format which have pixel set to 64:45 while for example StaxRip offers pixel at 16:11 for 16:9 movies. That results in final screen being a slightly bit more "wide". Not noticeable by eye but just it's better to know which shape model is the "correct".

7ekno

31st July 2011, 02:44

I prefereably like to use the aspect ratio that gives true 16/9 shape on monitor.

Then get out an old fashioned ruler :P

Find a resolution that gives 16:9 on the actual panel, then you can calculate the combined encoded and panel manufacturer pixel shape (cause at the end of the day, a 24 inch panel that is native 1920 x 1080 will have a different pixel shape to a 24 inch panel that is native 1920 x 1200) ;)

Tek

Anakunda

31st July 2011, 06:11

Then get out an old fashioned ruler :P

Tek

That's not good enough possible :D as I wrote I havenot precise enough ruler to find the difference ane beside this I don't know if my monitor is well calibrated (the H/V size can be easily changed by picture controls).
I ask differently then, in spite I can`t find the information of pixel shape on DVD cover, is there a list of publishers and pixel shapes they use for their DVD's? Or simplier, what of the three models is most commonly used, or most close to general standard.

Sharc

1st August 2011, 11:27

I ask differently then, in spite I can`t find the information of pixel shape on DVD cover, is there a list of publishers and pixel shapes they use for their DVD's?
Not that I know.

2Bdecided

1st August 2011, 12:29

cause at the end of the day, a 24 inch panel that is native 1920 x 1080 will have a different pixel shape to a 24 inch panel that is native 1920 x 1200?! both will almost certainly have square pixels, but 1920x1080 is a 16:9 display, while 1920x1200 is a 16:10 display.

Cheers,
David.

hello_hello

1st August 2011, 15:25

Or simplier, what of the three models is most commonly used, or most close to general standard.

Just use straight 16:9 or 4:3 resizing. If you're using an encoder GUI which gives you an option between the "official" ITU resize method and straight 16:9 or 4:3 resizing, use the latter. Most DVDs do (at least I'm sure they do) and most software players also use that method when playing DVD video.

If you want exactly 16:9 and you use the ITU resize method, which isn't actually 16:9 but a little wider, you'll have to crop a few pixels from each side of the picture to take you back to 16:9.
Which encoder are you using and are your encodes aimed for playback on a particular device? If possible, it's better to use anamorphic x264 encoding, which basically means you're using the same shaped pixels as the original DVD. You still have to pick the resizing method to use, but if later you decide it's the wrong one it should be fairly easy to change the encoded file's aspect ratio without having to re-encode the video.

Anakunda

1st August 2011, 15:34

If you're using an encoder GUI which gives you an option between the "official" ITU resize method and straight 16:9 or 4:3 resizing, use the latter.

Do you mean 64:45? If I want to preserve full matrix 720x576, I don't want to use anamorphic encoding but pass to encoder this source unstretched with --sar 16:11 or --sar 64:45. Since StaxRip offers by default the former I am a bit confused which is THE ratio I should use.

TheSkiller

1st August 2011, 22:02

Most DVDs do (at least I'm sure they do) and most software players also use that method when playing DVD video.
Newer DVDs (at least PAL ones) probably mostly use generic. Software players don't calculate the stetch factor by using a PAR, they just use whatever the flag (4:3 or 16:9) says (generic PAR).

However since DVD is derived from analog video, playing a DVD on a hardware player through an analog connection (Composite, S-Video, RGB, YPbPr in 576i) will always give you an ITU shaped picture as the image is being transformed to analog using 13.5 Mhz which assumes the inner 702 pixels of the 720 total pixels are the active picture carrying the 16:9 or 4:3 content. (Edit: PAL only, NTSC has different numbers)
All DVD players with HDMI that I've ever seen were using the ITU method as well to resize the DVDs picture to the requested HD square pixel resolution. That means they do something like
(512/351)x720/576=1050/576.
I've verified this with test cards. You can tell easily if you make a DVD with any kind of still picture (preferably a test card) and encode it once to 704x576 and once to 720x576. If your player is not using ITU then both will look exactly the same, as it is simply resizing both to what the flag says (4:3 or 16:9). From what I've seen I doubt that there are more than a hand full of hardware players that do not obey ITU via HDMI or YPbPr upscaling. The problem is just that there are many DVDs not obeying ITU.

64:45 is the so called "generic PAR". However you can't generally say "use the latter (generic)" as no one will garantee you that it is the right one for a given DVD. Instead you could just flip a coin and let it decide which one to use if you're not willig to try and check which one is the correct one for a given DVD. This can be tricky though, you need to find an object in the movie of which you can be sure that it has to be a perfect round object (someting like the moon, a ball etc).

It really is a shame that it is neccessary to find out, if DVDs were all made like they are supposed to when the spec was created (following the ITU) then it would be so easy! :mad: Authoring studios are lazy and/or don't care about the issue of sampling the picture correct in the first place.

Sharc

1st August 2011, 23:08

.....The problem is just that there are many DVDs not obeying ITU.....
Exactly.
Or often the active picture of 702x576 (i.e. with ITU pixel AR) is padded with black borders to 720 width, means a player which does not stretch according to ITU but sticks to 16:9 (1024/576) instead will display a slightly too narrow active picture.....

hello_hello

2nd August 2011, 08:23

64:45 is the so called "generic PAR". However you can't generally say "use the latter (generic)" as no one will garantee you that it is the right one for a given DVD.

I never guaranteed it would always be right, just that "most DVDs" use straight 16:9 resizing.... in answer to Anakunda's question as to which resize method is most commonly used.

hello_hello

2nd August 2011, 08:25

Exactly.
Or often the active picture of 702x576 (i.e. with ITU pixel AR) is padded with black borders to 720 width, means a player which does not stretch according to ITU but sticks to 16:9 (1024/576) instead will display a slightly too narrow active picture.....

Personally, I think most DVDs with padded black borders still use straight 16:9 resizing.

hello_hello

2nd August 2011, 08:47

Do you mean 64:45? If I want to preserve full matrix 720x576, I don't want to use anamorphic encoding but pass to encoder this source unstretched with --sar 16:11 or --sar 64:45. Since StaxRip offers by default the former I am a bit confused which is THE ratio I should use.

Anamorphic encoding is passing to the encoder the source unstretched. I assumed from your original post you weren't using anamorphic encoding because you were asking about resizing to dimensions such as 1024x576.

Whether you use anamorphic encoding (720x576) or you resize up (1024x576) or resize down (704x400) etc. and encode using square pixels, you still have to choose which resizing method you want to use.
Even though StaxRip defaults to ITU resizing, as does MeGUI, I'm of the opinion most DVDs use generic 64:45 resizing. I always use the generic method. Well.... once I did think the ITU resize method looked more correct, but aside from that one time I've never encoded a DVD using ITU resizing. Not even when encoding to AVI with AutoGK (which has ITU resizing disabled by default).

I'm basing my opinion on the numerous DVD encodes I've compared to the BluRay version and only once or twice have I compared the DVD encode and thought, "whoops, I should have ITU resized that one", and I've compared quite a lot of DVD encodes to their BluRay versions. Pretty much all the time, they have the same picture aspect ratio.

I think Terminator 2 was one of those "should have used ITU" encodes, but I may not be remembering correctly.

As I said too, when you're using anamorphic encoding picking the "correct" resize method when you encode isn't as important as it is when encoding using square pixels, because if you change your mind later on it's fairly easy to remux the MKV/MP4 and change the aspect ratio in the process. So if you encode an anamorphic DVD using the generic resize method and later on decide ITU would have looked better, it's an easy fix.

Anakunda

2nd August 2011, 09:07

2Bdecided

2nd August 2011, 15:10

I understand that the differences are behind eye's recognitionThey're not - a 2% difference is easily seen. What's not easy is figuring out which is correct without a reference - but that's quite different from saying that the difference is invisible!

Cheers,
David.

2Bdecided

2nd August 2011, 15:15

Personally, I think most DVDs with padded black borders still use straight 16:9 resizing.For that, they would have to be originated using a process that assumed 720 pixels defined the aspect ratio ("generic PAR"), then put through a process that assumed ITU PAR and cropped down to analogue PAL active picture while retaining 720 pixels (or another process that added black pixels back).

It's not impossible - a simple analogue dub will do this - but IMO it's more common that any digital video you find with ~8 pixels black bars either side was originated by something that respects ITU PAR (and analogue blanking)...

...simply because it's more common to start with an ITU-compliant camera (one way of getting black bars) than it is to start with some generic device and then put the video through some analogue connection (the other main way of getting black bars).

Cheers,
David.

TheSkiller

2nd August 2011, 15:48

I agree with 2Bdecided, a error of 2.5% can be seen. In fact I remember spotting that a DVD was using generic PAR while watching it (on my TV) when something appeared on the screen that was supposed to be a perfect circle but instead it was a slighly "fat" circle (wider to the left/right). I wasn't even trying to spot it, but it was just too evident the circle was not a circle.

I drew two circles in Paint, one is a perfect 240x240 circle, the other has an error of 2.5% which makes it 246x240.

http://img855.imageshack.us/img855/3677/errorpresentation.png

If you look just at one at a time (try to cover the other one with your hand), then it is very easy to see.

It is likely that the video has originated from something that respects ITU if there are borders on both sides of at least 8px.

smok3

2nd August 2011, 15:55

slightly unrelated, but i'am adding some code to my old calc (the code is about "human DAR" representations), the table i have so far is:

2.35 = old cinemascope
1.85 = US theater - since 1960
2.39 = US theater - current / Panavision / Scope
1.778 = 16:9
1.333 = 4:3
2.333 = 21:9 - philips tv
1.66 = 15:9 - Super 16
1.60 = 16:10 - LCD common
1.37 = Academy - close to 4:3
2.55:1 = wide old cinemascope
2.76:1 = Ultra Panavision 70 - Ben Hur

anything else/other important formats? suggestions? good/bad? do you have better numbers, ect? better name than "human DAR"?

edit: beta http://resizecalc.sourceforge.net/beta15/

edit2: to add:

Ghitulescu

2nd August 2011, 17:18

try 1.66:1

manono

3rd August 2011, 04:42

There are quite a few movies in 2.55:1 and some even at 2.76:1 (Ben Hur).

Ghitulescu

3rd August 2011, 08:01

How the west was won is also a very wide movie ...

smok3

3rd August 2011, 12:18

some ineteresting/funny reading
http://www.widescreenmuseum.com/widescreen/cscope-ac.htm
"stereo sound will give the illusion of 3d, ect"
or from http://www.widescreenmuseum.com/widescreen/superpanatar-c.htm
"Since all anamorphotic systems require the use of considerably larger screens, it is very important to choose a screen of high reflectance. Panavision, Inc., has found that the new lenticulated screens, such as the Magnaglo Astrolite, provide an exceedingly brilliant picture and distribute the light evenly to every seat in the theater."
and
"It must be well understood that the aspect ratio of any anamorphotic lens varies according to the focal length of the lens on the projector. In other words, an anamorphotic lens will project a slightly wider picture when used with a 2 3 /4" projection lens, than when it is used with a 6" projection lens. For this reason, the SUPER PANATAR is not supplied with a complete aspect ratio scale, and it is necessary for the projectionist to indicate proper index marks on the aspect ratio dial. "
ben hur
http://www.widescreenmuseum.com/special/ben-hur.htm

manono

3rd August 2011, 13:36

Hehe, I own one of those ridiculous letterboxed 2.76:1 aspect ratio Ben Hur DVDs. Warner Home Video is usually careful to get the intended aspect ratio correct so I'm not sure I accept what the guy at that link says. Even IMDB says it was created at 2.76:1, although it was usually shown at 2.5:1 or 2.55:1. The impression I got was that those less wide versions were created by cropping from the sides, rather than the 2.76:1 version being created by cropping from the top and bottom and leaving out vital picture information. In the DVD Beaver comparison, earlier 2.76:1 DVD versions did crop from the top and bottom (and from the sides also), but the later 4 DVD 2.76:1 versions (one of which I own) contained more information:

http://www.dvdbeaver.com/film/DVDCompare/ben-hur.htm

Although it does look a bit narrow on a widesreeen TV set, I'm glad to have it in that ratio.

Lyle_JP

5th August 2011, 23:23

You missed a big one in your aspect ratio list. 2.20:1, the aspect ratio of films shot on 5-perf height 65mm film with spherical lenses. Basically, nearly everything shot as "Todd-AO" or "Super Panavision". 2001, West Side Story, Lawrence of Arabia, Far and Away, Kenneth Brannagh's Hamlet, the original Tron, South Pacific, Around the World in 80 Days, and the Todd-AO version of Oklahoma* are just a few examples. This is often referred to as movies shot "in 70mm", which is not accurate as 70mm is an exhibition format, not a camera negative size (the additional 5mm is for multiple magnetic soundtracks).

*Every scene in Oklahoma was shot (at least) twice; once in Todd-AO, and once in 35mm CinemaScope (2.55:1). The reason for this is because Oklahoma was a rare example of Todd-AO which was filmed and projected at 30fps, not 24. Because of the frame-rate difference, the film could not simply be reduced to 35mm like most later Todd-AO features.

smok3

7th August 2011, 13:07

ok, will add. the name?

2.20 = 5-perf height 65mm ?

is this the aspect of the actual projection, or just the recording one?

Vern Dias

11th August 2011, 20:19

It's the actual projection AR. The recording AR for 65mm negative is actually closer to 2.35:1. 70mm mag stock has two mag tracks outside the perfs on each side accounting for the extra 5mm width and 1 mag track inside the perfs on each side accounting for the reduction to in projection AR to 2.20:1.

Vern

Lyle_JP

11th August 2011, 23:18

ok, will add. the name?

2.20 = 5-perf height 65mm ?

is this the aspect of the actual projection, or just the recording one?

Just call it "flat 70mm". :) Or "Todd-AO/Super Panavision". For the purposes of your list, it may be easier to refer to the exhibition formats, rather than the camera negative formats. Flat refers to the fact that there is no anamorphic squeeze. Ultra Panavision (also known as MGM Camera 65) was 70mm projected through an anamorphic lens producing a 2.76:1 AR, which I see is already on your list.

Technically, Todd-AO and Super Panavision are not identical formats, as Todd-AO lenses had a mild "fish-eye" distortion and were meant to be shown on deeply curved screens. Super Panavision was for flat screens. But both are 2.20:1 on home video.

Lyle_JP

12th August 2011, 00:08

A few more suggestions about your list:

1.66:1 is not just the AR of super 16, it has also been the standard "matted widescreen" format used throughout Europe for 35mm. It's their 1.85:1, if you will. The British (who must reflexively be different than the French at all costs) used 1.75:1 for their matted widescreen. Lately, 1.85:1 is becoming a more common worldwide standard now, supplanting both.

2.55:1 is not really older than 2.35:1 for CinemaScope; both CinemaScope ARs were common at the beginning. 2.55:1 is CinemaScope with Magnetic soundtracks, and 2.35:1 is CinemaScope with an optical soundtrack. That's why only CinemaScope films released in stereo were 2.55:1 (stereo sound would not come to optical tracks until the late 70s). 2.35:1 was also the proper AR for Panavision until the early 70's, when it was changed to 2.39:1.

Then of course, there is the silent film AR of 1.33:1 (which often ran at only 16fps), and the IMAX AR of 1.44:1.

Finally, there is Cinerama. In the beginning, Cinerama was a genuine film format, which used three synchronized projectors, and had an AR of 2.89:1! However, in most Cinerama theaters this was often cropped to anywhere from 2.59:1 to 2.76:1. In later years, Cinerama became more of a theater "certification", much like THX today, and films shot in Todd-AO or Super Panavision would be advertised as being "shown in Cinerama" if exhibited in a Cinerama theater. Apart from a handful of travelogues, only two narrative films were produced in genuine 3-strip Cinerama: How the West Was Won, and The Wonderful World of the Brothers Grimm.

I would create two lists. First, film exhibition aspect ratios:

1.33 - Silent
1.37 - Academy
1.44 - IMAX (full frame)
1.66 - European Flat
1.75 - British Flat
1.85 - Domestic (US) Flat
2.20 - Todd-AO/Super Panavision (70mm)
2.35 - CinemaScope (opt)/Panavision (old)
2.39 - Panavision (modern)
2.55 - CinemaScope (mag)
2.76 - Ultra Panavision/MGM Camera 65 (70mm)
2.89 - Cinerama

Then, video display aspect ratios:

1.33 - NTSC/PAL (4:3)
1.60 - Common LCD
1.78 - HD (16:9)
2.33 - Philips Wide

There's one I left off the list, because it is somewhat controversial. 2.00:1, also known as Univisium (http://en.wikipedia.org/wiki/Univisium), is the brainchild of Vittorio Storaro, who believes he has created the perfect filming and exhibition format that could replace all other aspect ratios both in the cinema and at home going forward. He loves Univisium so much that he started demanding that films he had photographed previous to 1998 be cropped to 2.00:1 on home video to help promote it. That's why we've been stuck with cropped widescreen versions of The Last Emperor, Reds, and (until very recently) Apocalypse Now.

For the record, Univisium has never actually been used in a movie theater, to my knowledge anyway.

blis

13th August 2011, 08:12

This is my first post and perhaps the most bewildering issue for a computing person to adapt to when entering the media industry... "the square pixel" versus the what?
My world of IT teaches me that a pixel is a pixel, it holds a numeric value representing the RGB and that's it. For me 16:9 is a ratio and that's maths. For PAL 16:9 I use 1024x576 (576/9*16) and all is good.

As for NTSC... it is evil!! 853.33333333 a third of a pixel.. really?

Sharc

13th August 2011, 09:17

....My world of IT teaches me that a pixel is a pixel, it holds a numeric value representing the RGB and that's it. For me 16:9 is a ratio and that's maths....
I would say you address a 'sample' rather than a 'pixel' - but often the terms are used for the same. The sample indeed has no shape but represents colour plus brightness. The "pixel shape" or "pixel aspect ratio" is a means to describe the horizontal vs vertical spacing of the samples in a 2-dimensional matrix arrangement.
For digital display devices the "pixel" (picture element) represents the basic area (square or rectangle) which visualizes a sample.

hello_hello

15th August 2011, 02:08

OK I think this is most explaining. I understand that the differences are behind eye's recognition, but for better feeling it's good to know that encoding with good params. Just for assuring StaxRip defaults regarding the table posted by Sharc MPEG4 shapes.

I encoded a Bluray disc today and as a result I opened my old DVD, AVI encode of the same movie to see what the quality was like. Before I even looked at the Bluray disc... as soon as I opened the AVI I thought "that should have been resized ITU", and compared to the Bluray disc obviously it should have been.

I don't know why it was so obvious to me today when three years ago I didn't notice.... but out of the hundreds of AVIs I've encoded, I think that's now four I've found which in hindsight should have been resized using the ITU method.

blis

20th August 2011, 02:04

I would say you address a 'sample' rather than a 'pixel' - but often the terms are used for the same. The sample indeed has no shape but represents colour plus brightness. The "pixel shape" or "pixel aspect ratio" is a means to describe the horizontal vs vertical spacing of the samples in a 2-dimensional matrix arrangement.
For digital display devices the "pixel" (picture element) represents the basic area (square or rectangle) which visualizes a sample.

Thanks Sharc. that's a good way of putting it :)

Gavino

21st August 2011, 10:56

My world of IT teaches me that a pixel is a pixel, it holds a numeric value representing the RGB and that's it.
Entirely agree - I wince when I hear talk of 'square' and 'non-square' pixels.
The term 'Pixel Aspect Ratio' should really be called 'Pixel Spacing Ratio', but the battle has been lost a long time ago.

smok3

21st August 2011, 20:05

i'd say that pixel is just fine, from wikipedia:
In digital imaging, a pixel, or pel[1], (picture element[2]) is a single point in a raster image, or the smallest addressable screen element in a display device; it is the smallest unit of picture that can be represented or controlled.
so PAR follows the 1st part of the definition, which is abstract enough to be represented/imagined by any aspect. So what exactly is wrong with "non-square" pixel? ("non-square representation of pixel" would be better?)

SeeMoreDigital

21st August 2011, 20:16

i'd say that pixel is just fine, from wikipedia:

so PAR follows the 1st part of the definition, which is abstract enough to be represented/imagined by any aspect. So what exactly is wrong with "non-square" pixel? ("non-square representation of pixel" would be better?)I'm with you ;)

hello_hello

21st August 2011, 20:42

Me too, I'll confess. Maybe it's because I've been "raised" to think in terms of pixels and pixel aspect ratios, but I can understand the concept of a display's pixels, and I can relate them to pixels in an image, even if those pixels aren't square.
Calling them pixels on a display while referring to them as samples in a video, and to their "spacing" if they're not "square" just seems to add an extra layer of confusion for the sake of semantics.

X264 uses sample aspect ratio rather than pixel aspect ratio doesn't it? I think I remember that confusing me at one stage.

Gavino

21st August 2011, 21:04

So what exactly is wrong with "non-square" pixel?
As the definition you quoted says, "a pixel is a single point".

Pixels (as represented in a digital storage medium) don't have a width and height - they are point samples. So Pixel Spacing Ratio would be a more accurate term than PAR. The distinction may seem academic, but it causes immense confusion when trying to understand what really goes on during resizing (ie resampling to a different sampling grid).

Pixels on a display do have an actual size, but they are rarely square (or rectangular).

See Alvy Ray Smith's classic 1995 paper: A Pixel Is Not A Little Square (http://alvyray.com/Memos/CG/Microsoft/6_pixel.pdf).

But I accept I am in a minority on this issue. :(

hello_hello

21st August 2011, 21:51

I couldn't get your link to work.

Gavino

21st August 2011, 22:27

Sorry, I copied it from an out-of-date reference, and it seems it has changed. I've fixed it now.

mpucoder

22nd August 2011, 04:52

Argue all you want about pixels being points or rectangles, but I'd like to see you understand subpixel resizing, rotation, anti-aliasing, deformation, or convolution using points. To most people dealing with digital graphics a sample is what a sensor sees, and it probably has some area (since a point has no area how would a photon interact with it?), but how much is not important. The sensor need not fill the space between sample points or have the same geometry as the sample points. For that matter it need not be rectangular. So, yes, in a sensor we are taking samples, and they are considered to be points in a grid.
Pixels, on the other hand, are what we manipulate mathematically, and they have a definite width:height ratio. During resizing, rotating, or deforming new pixels are formed by suming the products of the source pixels' value * area covered by the new pixel.

Gavino

22nd August 2011, 11:17

Argue all you want about pixels being points or rectangles, but I'd like to see you understand subpixel resizing, rotation, anti-aliasing, deformation, or convolution using points.
...
Pixels, on the other hand, are what we manipulate mathematically, and they have a definite width:height ratio. During resizing, rotating, or deforming new pixels are formed by suming the products of the source pixels' value * area covered by the new pixel.
On the contrary, and this is exactly my point.
All software (and research papers etc) that I have ever seen treat pixels as pure points on a grid.
Resizing, deformation, etc, is done based on discrete sampling theory (as an idealisation, of course), which treats pixels as samples of a continuously varying signal which can be reconstructed at any given point using the Nyquist-Shannon sampling theorem (http://en.wikipedia.org/wiki/Nyquist%E2%80%93Shannon_sampling_theorem).

smok3

6th September 2011, 21:43

kinda interesting, but really to call it a point (or vertex) is just a mathematical approach (this vertices are never rendered/displayed, its just a data container...). In practice they do get a certain aspect, which is easily described as rectangle or square for the sole purpose of describing the aspect of it, no?

p.s.
I do actually understand that stuff a bit (been working with 3d on and off for some years now), so yes vertex is a point (but that usually renders as black), so to see something you need to either (in case of lightwave 3d):
1. tell the software to render those as lil dots
2. make the geometry of it (at least 3 of this guys are needed)
3. apply some sort of material to those for the purpose of simulating foam/water/smoke, ect

Ghitulescu

9th September 2011, 12:45

In analog world, there are no pixels, there are samples. These are values sampled according to an algorithm. Example: digitizing a VHS.
For capturing devices (with CCD or similar eg CMOS), there are pixels, each photo-sensor will generate a pixel. It may be treated as such, interpolated to give a video signal, or RGB pixels (eg Bayer), but the information is no longer a sample, it is a pixel, it has a surface.
For displaying the things are much simpler, all are pixels in digital realm. The panel driver splits the signal into exactly the same number of values as display pixels are, and each individual pixel has a surface that radiates the same light. It is not possible to display 2 samples with a single pixel.