Academy ratio question (1 Viewer)

[Jeff Wilson]

Bear with me here if this seems crazy; In a discussion on my board, I am being told that DVDs of films shot in Academy ratio are being somehow "squeezed" at the sides in order to fit a television screen. The people I'm discussing this with claim this is a different thing from television overscan, which is hiding the outer edges of the image. Now, one of the people is capturing footage and running it through Premiere, which has some kind of function in which the picture is restored to what he claims is a true representation of the theatrical ratio. Is this nuts? Or am I missing something obvious? I mean, why would a DVD of a 4:3 movie on a typical 4:3 TV have to be squeezed in any way? They're more or less the same ratio, right (1.37 vs 1.33)? The original thread is linked below for anyone wishing further elaboration.
Aspect ratio discussion

 

[cafink]

He's full of baloney. He speaks as though he's knowledgeable about such things, but he thinks that televisions have an aspect ratio of 1:1.

 

[Peter Apruzzese]

While I agree with some of your posters about wanting certain films windowboxed, I've never heard of this so called "compression" problem. I would think that if this was a genuine issue, it would have been mentioned by now since home video is over 25 years old!

 

[Jeff Wilson]

Exactly. Windowboxing I can see, since I zoom out Academy ratio films to view on my DVD player, but the other part of it I just don't buy. It doesn't make sense.

 

[Seth Paxton]

First of all, this guy appears to be mixed up on a few items. He is operating on an MPG bitsteam in Premiere (for the Mac, right), a program I'm not familiar with. Even still, Premiere needs DATA to do anything with, and for some magic "maintain aspect ratio" change it would need a flag/field in the header of the MPG stream. I had no idea one existed.

Kane is stored on the DVD in a 1.33:1 pixel ratio. I cannot say for sure that when a transfer from FILM to any video source is done that some techique isn't used to squish the 1.37 film into a 1.33 space. However, it's never been mentioned anywhere that I have ever heard.

Also, viewing Kane on a CPU with the video sent for greater than 720 pixel width WILL SHOW the full picture that is on the DVD. Yes, TVs with overscan will lose the edges, but a computer will not. So there is nothing new or hidden in the image to somehow be found.

Back to that squishing thing. IF, if he is claiming that during a transfer that a 1.37 image is taking from the existing film cell and sampled in such a way as to result in a 1.33 pixel ratio, then he SORT OF is recovering the original image by capturing and stretching back to 1.37. Of course depending on how strong Premiere is at the pixel interpolation this may or may not be such a good thing.

NOW, if the transfer was done (as many P&S are, for example) by choosing a 1.33 square from the film cell image and just transferring that, then this guy is an idiot, period. Because that means that the second that transfer is done, long before he gets his hands on it, picture loss has already occured and the remaining image is in no way stretched/squished. And in that case trying to stretch out the image with Premiere is only going to damage the image by stretching image data that is in a correct geometric ratio.

THIS CAN VARY from transfer to transfer. Maybe the #1 easiest to access example off this going right now is RESEVOIR DOGS. 2 different DVDs, 2 different amounts of image taken from the original source. Both are still close to the same ratio (I think the new SE is slightly taller) but clearly not quite the same. There was no "compression" used here, they just took a larger section of the original image for the sample.

More does not always mean correct, otherwise these full frame open matte transfers would be correct, and we know they are not.

To follow up with that, this also means that unless he has access to original FILM PRINTS, not fucking video or some crap, then he has NO WAY to verify any "compression", nor to directly address the issue of how much of the source film image was intended to be seen.

I won't disagree that with overscan on so many sets, it would be nice if more films left room on BOTH the sides and top/bottom to account for that. But then again you are talking about not utilizing the full resolution on the SOURCE DVD/video, which makes no sense when the playback systems are quite varied and many are not hindered by overscan (and many more won't be in the future).


This is what tells me he is very confused...

these 2 little clicks took 4 hrs of rendering on 45 minutes of footage

Anyone want to tell me how a DVD player would make use of an Aspect Ratio field in the bitstream to "uncompress" video on the fly if this is the sort of rendering time we are talking about. Not exactly real time now is it?

Again, that's why there is no need for an AR field, simply because the devices couldn't make use of it.

There MIGHT (I stress MIGHT) be a INFO field telling the system what AR the pixels ARE IN, not what they are supposed to be in, and I question even that.

 

[Seth Paxton]

Here is an example of the "fix" he is proposing.
First the DVD image...

Then after I fixed it...

You can see how it looks too thin before, but after resizing it to it's 1.37 ratio the characters look correct rather than too thin.

 

[Seth Paxton]

Now to be honest with everyone.
What I just posted was not quite right. But I did want to make a point before I made this next post, just to get everyone in the right frame of mind.
What you see above is NOT the DVD image as it came off the disc. I CROPPED it to create a 1.37 image, then SQUISHED it back to 1.33 to artificially create a 1.33 image.
So the point was that the "good" or "fixed" image actually is the accurate portrayal of geometry of the image. The "original" represents an image that is, in fact, too thin or squished.
Let's do it again, but this time with the real DVD image and the "fixed" version. Keep in mind anything you thought you saw fixed before in the bogus comparison.
Real DVD image (1.33)

Stretched in Paint Shop Pro image (1.37)

Better, fixed? I don't know. Unless I see proof by means of comparison with the original film image or from the person who did the transfer, then I remain unimpressed.
And I still want to know where he got this magic "original aspect ratio" that Premiere is working with. In PSP the "maintain AR" would have kept the 1.33 ratio until I told it to switch to a 1.37 ratio. And on top of that, even if such compression is done, can we say that it is done equally across the image such that using Premiere does an unsquishing in the exact same manner?
I'm gonna guess a big fat NO WAY IN HELL.

 

[Jeff Wilson]

Thanks for the explanation, Seth. I had a general idea why this seemed bogus, but my tech knowledge isn't firm enough to describe it accurately. Judging from the captures you provided, even if one wanted to stretch a 1.33 image to 1.37, isn't the difference so minimal as to make it somewhat pointless? It is only .04 we're talking about here.

 

[Brian McHale]

I've wondered before about the difference between 1.33 and 1.37, but never bothered to check. My 42" widescreen Toshiba has a 4:3 viewing area equivalent to a 34" regular TV, which comes to 27.5 in(w) x 20.6 in(h). Multiplying the height by 1.33 gives a width of 27.398 inches, whereas 1.37 would be 28.222, a difference of 0.824 inches. I guess this would be noticeable, though not really significant.

Are these people implying that Academy Ratio 1.37:1 movies are being horizontally compressed to fit 1.33:1 TVs? I guess I always assumed they were cropped.

 

[Robert Harris]

1.33:1 is a silent aperture.

1.37:1 is Academy sound aperture.

Both lose information in projection via aperture plates.

My overall reaction to this is that someone has way too much time on their hands.

RAH