benefits of downsampling

Status
Not open for further replies.
T

tonacidigital

Well-known member
http://www.cornbread.org/FOTRCompare/index.html

The link above shows an example of how much downsampling can help. I like the "no uprezzing spoken here" motto. It'll be interesting to see how 4k footage downsampled to 1080p looks compared to natively shot 1080p. I'm sure the natively shot footage will look great, however, I plan on shooting 4k minimum...and perhaps 2k only when I need that 120fps.
 
:dankk2: Fun_tastic income.

This has been coming a large community of great quality members. IMHO, the world's best. Sorry my enthusiasm (just for the case of bothering you) but this works as a brotherwood, not a common forum where the people is posting today and next...year.

I'd like to know how does this presentation work? Take a discount on my self ignorance on some (just a few :laugh:) techie subjects.
 
tonacidigital said:
It'll be interesting to see how 4k footage downsampled to 1080p looks compared to natively shot 1080p. I'm sure the natively shot footage will look great, however, I plan on shooting 4k minimum...and perhaps 2k only when I need that 120fps.
Click to expand...

The idea of being able to shoot 4K "camera negative", with little regard what format you actually need is the big appeal for me.

I believe that the downsampled 4K to 2K or 1080 will be superior in "natural" sharpness and of course color sampling to the native 2K sensor solutions coming and even RED's own windowed 2K. I really don't need or want to think of the RED as a 4K camera, but as the ultimate-end-all 2K one... with a little extra when you need it for VFX, cropping, etc.

Personally, a 2K or 1080p post workflow is what I'll be doing, and have many reasons for thinking it's the "right" workflow for this camera.
 
Jim,

I believe the 4K editing can be useful for any late downsampling, being future proof. It's more a matter of that, than downrezing IMHO. But of course, it depends of your work, kind of medium and needs of each other.

Personally, a 4K post workflow (and downsampling route) is my bet, if possible -- that is, available for an affordable reasonable price.
 
Hi Emanuel, I agree completely with the idea of preserving the most possible resolution by shooting 4K, then using REDCINE to generate the highest quality 2K footage from that. I intend to shoot 4k for everything unless there's a frame rate or storage issue that would force me to choose another option. As you point out, you will always have the original 4K material to come back to if needed in the future for a re-conform.

At least for my work, I don't see much need for a 4K conform for projects in the near future. A 2K finish will be more than adequate for any e-cinema project I could imagine taking on. Having seen a number of 2K Hollywood projects projected digitally in the last couple years, I will be more than happy posting at that quality level, since the resolution on screen rivals any 35mm print I've seen after the various duplication stages the film negative has to go through to get to the cinema.

Until the time that 4K finish is commonplace, and venues for displaying your work in 4K are equally numerous, I will mostly use the RED as a super-sampling 2K camera.

As an aside, I also do dome show projects, which are well beyond a standard 4k finish workflow in scale, but that's another story and a situation the poor little resolution starved RED can't help with... :)
 
Hi Jim

Another advantage is the improvement in signal to noise ratio when you downsample.

A reason to go the 2K capture route though would be that S16 lenses tend to be much lighter for a given focal length.

Martin
 
2K or 1080 from 2.5K, 3K or ...

2K or 1080 from 2.5K, 3K or ...

This thread has made me think more about what Redcine can do and what the lenses we choose might enable. If you consider the sensor as just an array of pixels, going from 4K to a 2K window is based on current standards, etc. but is still somewhat arbitrary. With film, there are the actual limits of the size of the film standards, but this is all digital.

For instance many Super 16 lenses will cover, sometimes significantly, beyond the 2K window. I think this is the case for some of the S16 zooms (at least for part of their range) and I have also heard that the Kinoptic 16mm primes will cover well beyond the size of the 2K sensor window. Similarly, there are other lenses originally designed to cover 1" tube cameras that are used for Super 16 that would possibly provide even larger coverage. If you are shooting a scope sized frame, you might also be able to cover a larger area of the sensor before you see vignetting.

Now I realize for the sake of the Red team's sanity on the camera creation side, having a "choose your own format" might not be a very efficient way to go. I also realize that monitoring will be a little strange unless you find a creative way to mask off part of the frame at least on the LCD But humor me for a moment before you all tell me this is crazy, etc...

What if you were able to select the "window" in redcine (basically defininng how much you need to crop horizontally and vertically to avoid vignetting) that you wanted to have resized and output to 2K or 1080 or... Basically you would still shoot the camera in 4K mode - and of course take the storage hit and top frame rate limitations of 4K. Then you select your "window" in redcine if your lens can actually cover 2.5K or 3K or whatever. This way you get the benefits of that native resolution being downsampled to your post production format of choice, instead of just throwing that potential information away. And of course there could be some benefits for some slight reframing of shots as well, or even using this feature to reframe from a full 4K frame with 35mm lenses. And finally, for some who are using older cine lenses that aren't quite as sharp as new glass, this might allow us to get every scrap of "sharpness" we can from them by harnessing the benefits of downsampling.

Now I know some of you will say just use 35mm lenses and shoot 4K, but
if you're shooting with super 16 lenses or other lighter options to make hand holding easier or don't want the really shallow depth of field from the full s35 sensor area because you are working run and gun and don't have someone to worry about follow focus for you, it would be nice to have an option like this in Redcine. I believe the Arri D-20 actually shoots closer to 3K for 1080 output.

Red Team, any thoughts on this possibility? Or has this already been covered in another thread?
 
Some of OneCam's thought are repeats of previous threads though I don't believe anyone has connected these dots the same way. Someone on the Red Team, I think it was Graeme, confirmed cropping capabilities in RedCine. In my perfect world I could pull out whatever part of the full 4k frame I wanted, changing scales and center points on keyframes just like when I do "Ken Burns" style moves on hi res stills. Set the aspect ratio and manipulate a layover rectangle just like in the motion tab of Final Cut. Obviously the opportunity for motion tracking is huge. The folks that shoot on boats or other unstable platforms could use tracking points to create footage so smooth it wouldn't look real.

I know that for operators and DoPs the idea that an editor is going to reframe your shots is worrisome, or even vexing, but welcome to the new paradigm. A well defined and understood narrative with a good director and proper communication between the DoP and the post folks should avoid any serious carnage.

For my own work I like the concept that I can shoot handheld, no Steadicam to deal with, and then motion track any moves that look too radical (I step off a curb, the guy pulling on my belt to guide me as I walk backwards pulls a bit harder than I expect, I collide with the boom guy, etc). Yes, I understand the free floating style that a great Steadicam op can do has a place in the visual language that there is no true substitute for. I also know that for many projects I am likely to have a larger hand in, that budget constraints will make the day rate of top Steadicam guy difficult to wedge in.

In any case, I am a big believer in the benefits of downsampling. For most of my projects HD will be the delivery format, not 35mm blowups. If what I gathered from Mark Shubin is accurate, and he seems to know the physics pretty well, downsampling will create significantly more detailed images. There is a whole equation involving line pairs per millimeter of the lens and contrast ratios that I would not dare to paraphrase, but if you are skeptical you should google him - that's Shubin, not Cuban, though Mark Cuban may be a more entertaining google search :evil:

Blair S. Paulsen
RedOne #19
 
Status
Not open for further replies.
Back
Top