Welcome to our community

Be a part of something great, join today!

Register Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

What is 6K?

The way it is in my head, Mike, is that the debayering process results in 80% of the original recorded resolution. So if you shoot 5K and then ask for 5K output, the software takes that, debayers to 4k, and then upscales to 5K.

Whereas with 6k --> 4K, you start with 6K, debayer to around 4.8k, and then further downscale to 4K which makes the image even sharper and "better resolved".

Granted, all of this could be wrong, hence the root of my question.
 
Nick Pasquariello said:
So if you shoot 5K and then ask for 5K output, the software takes that, debayers to 4k, and then upscales to 5K.
Click to expand...

That was my initial (mis)understanding, but Graeme clarified in the other thread that this is not the case.

pixel resolution != measured resolution
 
Last edited:
Mike Nagel said:
That was my initial understanding, but Graeme clarified in the other thread that this is not the case.

pixel resolution != measured resolution
Click to expand...

My read on that is . . . how do I put this into words. You know how lenses have MTF ratings? How many line pairs of a given size that they can resolve? So in camera, you've got the individual photovoltaic cells taking a reading. There is a definitive limit to resolution based on how many cells there are. At 1080p you could have up to exactly 1080 lines (alternating black and white) across your image. But if the measurements the cells are taking have to be combined together by algorithm to create a digital image, then you couldn't get 1080 lines out of a 1080 bayered sensor. If you're algorithm maxes out around 80% of your sensor resolution, then you could get a digital resolution of around 864 lines out of your 1080p physical bayered sensor.

In which case pixel resolution != measured resolution.

If you take that debayered 864 line image and output it in 1080p, then you are upscaling from 864 pixels across to 1080. You'll still have only 864 lines (your measured resolution) and you'll be viewing it at 1080p (your pixel resolution). In terms of the digital process, the image has been resolved down to 864p, and then upscaled up to the pixel resolution of 1080p.

Which is different than an algorithm that takes 1080 lines of bayered cells, and calculates, pixel by pixel, 1080 pixels across as it goes. Upon a second reading just now, this latter option seems to be what Graeme is indicating. Hmm. So this would mean 6k = 6k, but 6k monochrome by definition is 20% sharper than 6k Bayered. 4.8k monochrome would be about as sharp (ignoring lens differences) as 6K bayered. Though, I guess however it's doing it, this is gonna be the case.
 
Nick,

absolutely. And that's what I tried to outline in post #40.

Multiple ways of looking at what could bring the best final results.

The question is if it brings improved results to debayer to 80% of the captured pixel resolution to get a "superior" debayer and then go from there... (?)

That's was my initial thought (and I wanted to test that) but RCX does not have a 80% default scaling option for the R1 MX which shoots 4.5K... (although you can manually define any output resolution you want, but then again 80% is a "rough" number so what should you actually use to run this test ?)


I think we are hijacking Phil's thread... maybe best to discuss further somewhere else... :mellow:
 
I don't think it's that much of a hijack. I've had discussions with Producers/ACs/DPs/Editors/Interns about resolution. If you have a firm understanding of what is actually going on under the hood, then you have a huge advantage in leading people in the right direction. And in being confident in what you're saying. It may seem tangential or irrelevant, but I can guarantee you I'll have a discussion sometime within the next 3 months based around what resolution to shoot at and this exact part of this thread will come in handy.
 
I think since the days of the Epic it is less of a topic as you capture way over 4K... so u will downscale to 4K no matter what and always get the improvements of that...


But a great example for the theory behind it is the Scarlet @ 4K and I for one would like to know what results in superior final IQ:


(1) capture 4K > debayer to 4K - 1:1 pixel resolution debayer and have image with roughly 80% measured resolution

vs.

(2) capture 4K > debayer w/ included scale to 3.2K (0.8:1 pixel resolution and possibly improved debayer quality) > upscale to 4K
 
6K vs 2K

6K vs 2K

Sumo Fights

sumo.jpg

6K vs 2K
 
So to be fair when comparing 3 chip vs single chip, a three chip 1080p camera with 6.3 million pixels is equivalent resolution to what pixel count bayer sensor..?


Whilst 1080p v 6k may seem a moot point at the moment, 3 chip >2k cameras are probably on the horizon for TV broadcast, so 3 chip technology will re-emerge as an option for high fidelity >4k images.



Mike Brennan
 
Awesome, great job Phil. This pretty much tells the story.
 
  • Thread starter
  • Moderator
  • #54
Tom Lowe said:
Awesome, great job Phil. This pretty much tells the story.
Click to expand...

Thanks Tom. There's actually so many ways to use the 6K resolution when it comes down to it.

I find it funny how it appears to certain articles and blogs online how "difficult" 4K is for an independent production, but the one thing Red did very right in that arena was provide a more affordable option to produce theater quality results for all productions big and small. The more 1080p and 2K cameras I see out there and how prepared I am to shoot in 6K gives me this odd sensation of feeling like I'm from another planet. There's a camera out there for everybody, but for those who are focused on image quality it's going to be an easy choice with Dragon.

I'm looking forward to what you're shooting abroad man.
 
Michael J Brennan said:
So to be fair when comparing 3 chip vs single chip, a three chip 1080p camera with 6.3 million pixels is equivalent resolution to what pixel count bayer sensor..?

Whilst 1080p v 6k may seem a moot point at the moment, 3 chip >2k cameras are probably on the horizon for TV broadcast, so 3 chip technology will re-emerge as an option for high fidelity >4k images.

Mike Brennan
Click to expand...

There is no equivalent. a Bayer sensor captures only 1/3 of the data per pixel that a 3 chip does. No amount of pixel count or sensor size can make up for that.

But, the way I see it (please chime in), the only question you need to ask is: How accurate is the color of a CMOS sensor after debayer compared to a 3-chip sensor ?

and that obviously all depends on the accuracy of the guesswork of the debayer algorithm. CMOS has delivered stunning results over the years, that's why everybody is using it.

It can never be perfect, but if color error is below dE 3 you might not even notice (that's a general rule of thumb, on some colors you can visually notice dE of 2 or less)...

But keep in mind, in Film & TV almost all footage gets graded anyways, so (to some degree) some color errors are irrelevant as you will change the color anyways.
 
Michael J Brennan said:
So to be fair when comparing 3 chip vs single chip, a three chip 1080p camera with 6.3 million pixels is equivalent resolution to what pixel count bayer sensor..?
Click to expand...

I disagree slightly with Mike's reply. Based on the 80% rule, a single chip camera would have to be 2.5K to provide similar (luminance) resolution to a 3-chip 2K camera. I'd go to 3K for good measure. :-) I assume, perhaps incorrectly, that it's cheaper to build a 3K single chip camera than a 2K 3-chip camera. You lose some colour but save in cost and, perhaps, camera size.
 
Karim is absolutely right, measured resolution is another difference between single chip and 3-chip. Fortunately less of a concern with Red since you'd be down scaling from 4K|5K|6K to 2K|1080 to have a 1:1 pixel comparison with a 3-chip.
 
Michael J Brennan said:
To be pedantic, 3 chip HD cameras have 3 x 2.1 million pixel sensors for a total of 6.3m pixels of camera output. When output is 4:4:4 it is a 6.3 mp image of non bayered or "guessed" info.

3chip tech ain't dead yet, three chip 4k and 8k output TV cameras are a feasible option as an effective way to enable use of high zoom ratio 2/3 inch TV zoom lenses, just saying :) no one is saying 6k isn't sharp as tacks.
Click to expand...

While technically 3 chip cameras may have individual colour sensors each with for example 2.1 million pixels, and camera manufacturers will often market such a camera as then being a 6.3 MegaPixel (MP) camera, the resultant images from it are NOT 6.3 MP. At it's most basic, to make a single nonguessed "colour" pixel, three perfectly physically aligned mono pixels behind colour filters/dichroic filter/prisms are required. So the basic images from such a camera are only 2.1 million pixels in resolution. There are some 3-chip techniques like offsetting the green pixel array which can increase the useful effective resolution with some clever processing, but nowhere near doubling of resolution let alone a tripling.
 
Mike Nagel said:
There is no equivalent. a Bayer sensor captures only 1/3 of the data per pixel that a 3 chip does. No amount of pixel count or sensor size can make up for that.

...
Click to expand...

This statement is incorrect if it meant to say that a Bayer sensor with x physical pixels captures only 1/3 of the data of a 3-chip camera utilising 3 sensors of x physical pixels each. Without more information about all the hardware, software and signal processing for a particular camera, it is impossible to say one is "better" than the other.

Both single chip and multiple chip approaches are valid ways to capture colour images. Both have pros and cons, and usually the tradeoff of compression with artifacts. Which approach is appropriate depends on the particular camera manufacturer's chosen design parameters for a particular target market of users.

Talking only about real detail capture, the only general statement that can be made is that more pixels allows the potential to capture more measured resolution. As Graeme and others have argued, it's only measured resolution (without significant artifacting) which is useful for comparison. Physical pixel count as a way to compare the two approaches is of very limited value for real world images.
 
You must log in or register to reply here.
Back
Top