Squeeze .vs Letterbox - Technical Theorem

M

marjamar

Well-known member
Now that I FINALLY have the DVX100 ("A" no less  ;D ;D ;D ;D ;D), I'm hoping to get an answer from Stewart (Stuart ?) English from Panasonic (need his email first) on a question that is really starting my brain cells to vibrate ;).  Until I do, I'll pose the question here for any to try and answer or debate, or at least think about.

We all know that squeeze is the same as letterbox in quality, because both are using the same (roughly) amount of pixels either way, right?

Hold your horses a minute before you jump off that bridge... :-/

Just suppose, if the following where true -

True - Letterbox is exactly the same processing in-camera as 4:3 up to the point the top/bottom bars are overlaid.  This would happen before compression.

Maybe True - Squeeze only uses the area on the CCD's which would comprise the 16:9 image.  The remainder of the CCD's data is either disregarded or discarded at some point along the way.

Question - If both questions above are true, would it be possible (even logical) that during the processing of the squeeze mode image, the datastream (in the form of 720x360 pixels), is actually being uprezed to 720x480 BEFORE it gets to the codac?  If it is, give this a think:

Would not the process of a hardware uprez prior to compression (which most have agreed is constant in nature no matter what else happens in this camera) produce a higher quality compression based on the theorem that only the portions of this data we actually use will be compressed?  

Think about it - In letterbox or 4:3 mode 86,400 pixels in each frame are being compressed to only be thrown away in post.  Conversely, in squeeze mode those 86,400 pixels are actually part of the 16:9 image we use.  Someone care to do that math here?  I could, but I'm a bit lazy this morning and besides you guys can do something here - right ;)

Some test here by Barry Green have shown that uprezing is one way to enhance the inherent quality of DV when done using a quality process.  If this is true (and I'm sure it is), would it not also be true for a hardware in-camera uprez process?

If this theroem holds water, it might be necessary to rethink the usefulness of using the in-camera squeeze mode .vs letterbox or 4:3 when producing 16:9.  It could in-fact make the Pana anamorphic lens mostly (dare I say it...) obsolete!

Comments (non-personal ;D) by all means welcome.

-Rodger
 
Re: Squeeze .vs Letterbox - Technical Theorem

..no matter how the cam does the rescaling(which is done by taking the letterbox image and uprezzing). the end result is inferior to the anamorphic, and could be easliy done in post with the same , or even better results by upscaling the letterbox footage. it has been tested , and failed miserably against the hopes we had all had for it.
 
Re: Squeeze .vs Letterbox - Technical Theorem

According to Panasonic and dvfilm the best rez for film out is with the Anamorphic lens adapter. Although "November" was shot 4:3 and croped in post, go figure 8)
 
Re: Squeeze .vs Letterbox - Technical Theorem

I have viewed tons of footage shot with and without the ana adapter blown up and projected on 12-20' screens and seriously, I cannot visually see any difference quality wise between the two... seriously. Perhaps maybe there is a sweet point (maybe on a 100" TFT tv) that you can tell, but at that size it was neglegable, and if you are viewing it on a 36" tv, you wont tell.

No for your compression theory, it would make sense that the compression would save some info,
but the problem is that Mini DV25 compression is divided between units of 8 pixels, so every 8 pixel "square" is alocated a certain quota for compression and size, so if you save a few kb compressing a black 8 pixel box on the top of the frame, that savings will not be transfered to the middle image being able to use that saving in compression. Its a little hard to explain or understand.

A better way to look at it is by flipping the concept and looking at what the anamorphic does. It squeezes more than 1 pixel line vertically of resolution into 1 pixel line. Problem is that the resolution of the CCD is fixed at the resolution of the lines, so even though you are putting more info into that line, you are loosing pixel detail that you normally would have in a 1:1 ratio. you cant record it all because the pixel can only contain so much data. That is why the anamorphic lens theoretically should offer more of an advantage than it does.
 
Re: Squeeze .vs Letterbox - Technical Theorem

Well, Jarred was kind enough to give me an email address for Stuart English, and I sent an email with a condensed version of my question. Hopefully he or someone there will have a reply for me before too long.

Also, just to clarify. I am not suggesting that the anamorphic adapter would not be better. Even though I have see less then clear footage shot with it, given proper use (focus), it would (should) be better then in-camera squeeze mode. I only suggest that if my thinking is not totally faulted, the anamorphic adapter may not be "as useful" as some my believe is it (was). For certain, my hopes are that my theorem is true, as I will absolutely have need of a tele-converter and since both can't be used on the same shot, I really want in-camera squeeze to be what I hope it is ;D.

-Rodger
 
Re: Squeeze .vs Letterbox - Technical Theorem

[quote author=marjamar link=board=Events;num=1076431425;start=0#0 date=02/10/04 at 08:43:45]
True - Letterbox is exactly the same processing in-camera as 4:3 up to the point the top/bottom bars are overlaid.  This would happen before compression.
[/quote]

Assumed true.

Maybe True - Squeeze only uses the area on the CCD's which would comprise the 16:9 image.  The remainder of the CCD's data is either disregarded or discarded at some point along the way.
Click to expand...

Assumed true.

Question - If both questions above are true, would it be possible (even logical) that during the processing of the squeeze mode image, the datastream (in the form of 720x360 pixels), is actually being uprezed to 720x480 BEFORE it gets to the codac?
Click to expand...

Unquestionably true.

Would not the process of a hardware uprez prior to compression (which most have agreed is constant in nature no matter what else happens in this camera) produce a higher quality compression based on the theorem that only the portions of this data we actually use will be compressed?  
Click to expand...
Assumed true.

If this theroem holds water, it might be necessary to rethink the usefulness of using the in-camera squeeze mode .vs letterbox or 4:3 when producing 16:9.
Click to expand...

Well, now we may disagree, or maybe not. Adam Wilt shot some rez charts with the DVX100A. In interlaced mode, squeeze mode was actually on par with, perhaps even inferior to, the PD150's in-camera 16:9. That was a stunning defeat -- the PD150 has lame 16:9 in-camera, so it was much disappointing to see the DVX perform on par with it.

In progressive mode, however, the DVX/A squeeze mode delivered about 360 lines of resolution, which is quite excellent.

I would definitely recommend using squeeze mode over stretching in post. It's possible a post-stretching algorithm will do a better job than the in-camera (because the in-camera MUST execute in real-time, but in post it can take as long as it needs to, to do a good job) but the compression issue does exist, and besides, in-camera squeeze in progressive/thin mode does a great job.


 It could in-fact make the Pana anamorphic lens mostly (dare I say it...) obsolete!
Click to expand...

No, here you're off by a country mile. While in-camera squeeze mode, in progressive/thin mode, is going to deliver the highest resolution of any prosumer 16:9 camera, at 360 lines max, it still doesn't hold a birthday-cake candle to what the anamorphic adapter can do.

Squeeze has enormous advantages, which I've listed elsewhere but will try to remember here: squeeze mode is free, it allows you to use screw-on filters, it allows you to use wide-angle and telephoto adapters, it allows you closer focus, it allows sharp full-telephoto focus, it allows you to use matte boxes, and it allows you to use the mini35. But it's also maxxed out at 360 lines, whereas the anamorphic adapter can deliver a full 480. So it's not the highest-quality option. But I'll bet it will be a compelling option for most shooters.
 
Re: Squeeze .vs Letterbox - Technical Theorem

There is a reason some people dont use it, Personally I would probally never use it again. Besides the thing being Plastic, the mounting of the unit is very inaccurate and cna create a quazy-back focus problem... The distance for the lens is so critical even a fraction of a mm can make a difference, and give a swing tilt effect (which I guess can be good).  You also need to manually vertically align the lens (rotate it till its "horizontal") which means if its not accurate you are creating a horizontal stretch as well as a vertical stretch.

That said, we have taken some beautiful footage with the lens (art history was shot entirely with the anamorphic lens, even though the body did get cracked on day 1)

Now when Century finally releases thier version of the anamorphic adapter, that will be the way to go (if you can afford it)
 
Re: Squeeze .vs Letterbox - Technical Theorem

as an owner of the dvx1oo with ana lense, i havnt given much thought to the 100a , but now that "november was shot on the 100 stock and cropped in post, ive given some little consideration to trading the 2 in for a new A model.. simply because its easier to shoot without the anamorphic, but again why not keep it and use it when i feel like it ,and when i dont, i can always just crop and strech in post just like "november"?..well see ill have to give it some more thought.
 
Re: Squeeze .vs Letterbox - Technical Theorem

Hey Jarred,

I understand the DCT formula for DV compression pretty well, but I'm not suggesting any alteration to any algorithms used here. Only that by (in effect) pre-cropping each frame before it is compressed, then using hardware scaling (uprezing) the cropped image back to 720x480, all before compression occurs, a higher quality to the finished frame would result in comparison to 4:3 or letterbox cropped in post. Nothing is thrown away after compression, which also means everything compressed is used - Make a choice, 0% loss after compression or 25% loss after compression. Give me the 0% loss ;D

Much of my thinking in this regard is spawned by my work in MPEG (all versions) for the past 10 years or so. Prioritization of the encode process in specific for final output can very greatly change how well your MPEG's will look in each format.

I really can't see any other way or reason this camear's squeeze mode would logically be processed.

-Rodger
 
Re: Squeeze .vs Letterbox - Technical Theorem

hey guys, im a shooter with limited post production experience. i mean, i know how to cut a scene very effectively... but my technical chops are a little behind. could someone explain to me the process of the "crop and stretch" that "November" used?

thanks so much.

daniel
 
Re: Squeeze .vs Letterbox - Technical Theorem

..it means you simply crop your 4:3 material to the aspect ratio you want..ie:16:9 or 2.35:1 and then stretch that image, or "blow it up" it, to fill a full d1 resolution window at that given aspect ratio.
 
Re: Squeeze .vs Letterbox - Technical Theorem

Hello everyone,

Got a reply from Jan Crittenden (new product manager for the DVX100 at Panasonic) for the email I sent to Stuart English. Her reply was specific to whether or not squeeze mode would produce higher quality then either 4:3 or letterbox cropped in post. Although she said "no", she continued to qualify that saying "at least not much". She also confirmed that only the portions of the CCD's are used for 16:9 (squeeze mode), which we knew, but also stated we would see a difference in quality on a "high quality monitor".

For the most part, this is exactly what I see. I have viewed a numbers of full quality clips from the anamorphic on my HD quality monitor and now comparing those to my own footage shot with the DVX100A in squeeze mode, the only difference I see is:

1.) No barrel distortion.
2.) No out of focus at extrema zoom.
3.) No loss in visible quality.

So, I still think my reason for this post is valid. Maybe not validated, but still valid. My hopes are along the way someone who has the DVX100A and the Panasonic Anamorphic Adapter can do some high quality comparison shots for us to view. If so, I can host them (up to about 4 GB per day anyway). So, anyone have the hardware and interest?

Remember, this new version of the DVX100A is really just out. Many are thinking it is absolutely necessary to get an anamorphic adapter to do quality 16:9. If this *is* true then fine, but if it isn't necessarily *true* it would be right for us to make it known.

-Rodger
 
Re: Squeeze .vs Letterbox - Technical Theorem

..again , the sundance cinematography winner was cropped and strecthed from the stock dvx, so it is obviosly possible to create "quality" widscreen images through either cropping and stretching, or usage of the squeeze mode on the dvx100A. really the only queston at hand is really whether the anamorphic adaptor offers a higher quality than the other methods. i have personally heard confilcting reports on the subject, but am definitley leaning towards the anamorphic since it is the method chosen by most other productions, and personally owning one, i know that it is capeable of clear and sharp images. and i can only assume that a considerable blowup would be noticeably better if coming from a higher resolution scource. espescially if cropping to 2.35:1 .
 
Re: Squeeze .vs Letterbox - Technical Theorem

thanks for the response, iamaloser. just to be clear, cuz i think i understand... the term "stretching" in this case, doesnt really mean "distorting" the picture size... but rather just blowing up that cropped portion of the picture to a "larger" size?

also, what is D1 resolution in comparison to miniDV?

thanks again.
 
Re: Squeeze .vs Letterbox - Technical Theorem

..pretty much yeah. your not ditorting the video, your cropping the video to a certain aspect ratio, and then enlarging the image.. the aspect ratio (proportions) stays the same..imagine your cropping your 4:3 720x480 d1 (roughly)video to a 16:9 window. you would have roughly a 360 x 480 window .. now you have to expand that to fill a full res 16:9 window which is roughly 864 x 480  so your cropping the image to a ratio, and then stretching it to fill the full res size as if you had shot with anamorphic to begin with. you could also crop to 2.35:1 and strech to a 2.35:1 window. but as you can probably guess the image will lose sharpness when expanding it like this. all pending on what sort of method /software/hardware is used to scale the image in size... you could just blow it up to high def if you wanted.. or whatever.  for example if you just load an empty  16:9 project in your nle , youll notice that the window is roughly 864x 480.. (you would use this for somthing shot in anamorphic..and the anamorphic video would unsqeeze to fill the full width).. now if you crop your 720x480 video to 16:9 youve got a 360x480 window, and you need to scale it up to fil the full res 16:9 window..  blah blah im repeating myslef as usual.. hpe i was helpful
 
Re: Squeeze .vs Letterbox - Technical Theorem

ahhh. i gotcha. its ok about repeating yourself, it drove it into my head. =) thanks again for your help.
 
Re: Squeeze .vs Letterbox - Technical Theorem

[quote author=XDeathOrGloryX link=board=Events;num=1076431425;start=0#13 date=02/11/04 at 11:54:55]
also, what is D1 resolution in comparison to miniDV?

[/quote]

D1 is 720 x 486, vs. DV at 720 x 480.
 
Re: Squeeze .vs Letterbox - Technical Theorem

Ok I'm not exactly following this completely here, this talk about stretching in post after you've cropped... Lets say I'm shooting a film without the anamorphic adapter, shooting in regular 4:3 mode with the original dvx100, and with tape i mark off the 1.85 framing on my monitor so that I can crop out to get a widescreen look in post... what need is there to stretch or do anything to my image... and why would it lose any resolution quality if all i'm doing is putting black bars over a tiny percent of the top and bottom? i'm confused here, it seems pretty cut and dry just shoot regular, mark off your frame on the monitor or lcd screen and mask in post.. now whats wrong with this? please help me understand all this talk.
 
Re: Squeeze .vs Letterbox - Technical Theorem

Hello John,

What your referring to is producing to "Letterbox" not 16:9. What this post was referring to was producing to 16:9.

To produce to 16:9 widescreen format, stretching is necessary anytime there are less then 720 x 480 pixels in the output format (NTSC DV format). Cropping (deleting) the 120 vertical pixels (60 pixels each from top and bottom usually) is necessary to produce the proper aspect ratio of 16:9. Then the remaining 360 vertical pixels will need to be resampled and stretch up to 480 pixels to be compliant to DV standards widescreen format. At this point it is also necessary to change the pixel aspect ratio (done automatically in editing software) to achieve a fully compliant DV widescreen format.

-Rodger
 
Re: Squeeze .vs Letterbox - Technical Theorem

im just going to reaffirm what marjamar said.

your right john b. , you can just crop your image to a 360x480 16:9 image and there ya go you have a 16:9 image. whic is great for the web,  but you cant broadcaast a 360x480 image, or even make a dvd out of it.  it needs to be in standard ntsc format as marjamar explained above. im sure you could do a film out from a 360x480 image, but you would be better off uprezzing it to widescreen dv (roughly 864x480) or high def before doing the film out, or projecting. because your dealing with more optical area for the film print machine,or projector. and since they work optically . increasing the physical optical resolution will have obvios advantages over having the same small digital copy. projecting the same digital image from a pinhead, or a 35mm size area will have different results. the 35mm size area will work much better. so blowing the physical area(digital dimensions) up to at least the standard is always a good idea  for really anything you plan on doing.. you can always make a smaller copy if you want it smaller for some reason.
 
You must log in or register to reply here.
Back
Top