PDA

View Full Version : 2K vs 1080p resolution question



Dominique Grenier
05-15-2007, 09:09 AM
There's something I'm not getting here. Considering that RAW is 1 channel and RGB 3 channels:

2K Raw
(2048 x 1080) * 1ch = 2211840 pixels, right?

1080p RGB
(1920 x 1080) * 3ch = 6220800 pixels, right?

So, there's something I must not do correctly, because with those calculs, 2K RAW is about 36% the resolution of 1080pRGB. Worst, its 80% the resolution of 720pRGB.

Now, if I'm right, outside of the obvious reasons to shoot RAW, what is the interest of 2k?

Edit: removed the MP info, since it seems it is not accurate, nor needed anyway...

TimothyD
05-15-2007, 09:12 AM
I'm no math whiz, but here is my layman's understanding...

With RAW you have three channels as well, so:

(2048 x 1080) * 1ch = 2211840 pixels = 2.21 MP x 3 = 6.62 MP

Kevin Halverson
05-15-2007, 09:28 AM
RAW is pre de-Bayer, so it is based upon a factor of 1 not 3. 2048 x 1080 = 2,211,840 pixels or 2.1 MP.

Also, keep in mind that 1,000,000 pixels is not a mega pixel, it is .953 mega pixels. The term mega is based upon base 2 math (2^20).

Chris Kenny
05-15-2007, 10:01 AM
So, there's something I must not do correctly, because with those calculs, 2K RAW is about 36% the resolution of 1080pRGB. Worst, its 80% the resolution of 720pRGB.


Multiplying out the number of color channels like that is not really meaningful. In acquisition, it is meaningful to talk about the fact that a Bayer-pattern sensor doesn't deliver as much actual resolution as a sensor which picks up all three color channels at every point. But in practice, it delivers probably better than 70% of the resolution, not the 33% that counting color channels would give you.

Anyway, the upshot here is, there probably will be more actual image detail in a 1080p image scaled from the 4K area of the sensor than in an image shot in 2K windowed mode. If you want the best possible 2K (or the best anything, really), the right approach is probably to shoot 4K RAW and scale to 2K in post.

Stuart English
05-15-2007, 10:24 AM
The answer to your dilemma is - RGB math and RAW math (and especially how these sensors work on a camera system) are not well understood. If you were to see the visual results then there would be no arguments from you, of that I am certain.

Brook Willard
05-15-2007, 11:18 AM
2048 x 1080

I think it's 2,048 x 1,152... not sure, though.

Dominique Grenier
05-15-2007, 11:32 AM
The answer to your dilemma is - RGB math and RAW math (and especially how these sensors work on a camera system) are not well understood. If you were to see the visual results then there would be no arguments from you, of that I am certain.

Well, I'm not really in a dilemna, since I intend to shoot in RECODE 4K RAW anyway... I'm working on a little software of my own to plan storage/backup needs, and other post-prod related management. And then I came to that conclusion I found somewhat odd...

So, if I understand correctly, you can't really compare the two based only on their pixels count, right? That would mean resolution is not just a matter of numbers but more a visual thing?

Joshua Provost
05-15-2007, 11:36 AM
There's something I'm not getting here. Considering that RAW is 1 channel and RGB 3 channels:

2K Raw
(2048 x 1080) * 1ch = 2211840 pixels = 2.21 MP, right?

1080p RGB
(1920 x 1080) * 3ch = 6220800 pixels = 6.22 MP, right?

So, there's something I must not do correctly, because with those calculs, 2K RAW is about 36% the resolution of 1080pRGB. Worst, its 80% the resolution of 720pRGB.

Now, if I'm right, outside of the obvious reasons to shoot RAW, what is the interest of 2k?

Adding the channels to find the resolution is flawed, but people make this mistake all the time when discussing 3CCD systems. The RGB channels overlay to create an image with resolution equal to a single sensor.

In this case, RGB implies that the image has been debayered and will be stored as RGB. Yet, regardless of how the image data is stored, it is a single CMOS sensor at play. It has a finite amount of pixels that no math can change.

Dominique Grenier
05-15-2007, 11:37 AM
Multiplying out the number of color channels like that is not really meaningful. In acquisition, it is meaningful to talk about the fact that a Bayer-pattern sensor doesn't deliver as much actual resolution as a sensor which picks up all three color channels at every point. But in practice, it delivers probably better than 70% of the resolution, not the 33% that counting color channels would give you.

So, the datarate of 1080p will be much higher than 2K RAW, but the visual aspect will be somewhat similar, or closer at least...?

JD Holloway
05-15-2007, 11:48 AM
It depends on if you downrez to 1080p from 4K, or do a 1080p windowed acquisition.

Downrezed 4K to 1080p will rock. Windowed 1080, not so much....

2K RAW windowed is better than 1080 windowed, and is a much lower data rate.

Michael Schrengohst
05-15-2007, 11:53 AM
There's something I'm not getting here. Considering that RAW is 1 channel and RGB 3 channels:

2K Raw
(2048 x 1080) * 1ch = 2211840 pixels, right?

1080p RGB
(1920 x 1080) * 3ch = 6220800 pixels, right?

So, there's something I must not do correctly, because with those calculs, 2K RAW is about 36% the resolution of 1080pRGB. Worst, its 80% the resolution of 720pRGB.

Now, if I'm right, outside of the obvious reasons to shoot RAW, what is the interest of 2k?

Edit: removed the MP info, since it seems it is not accurate, nor needed anyway...


A 2K Academy image is 2048 pixels wide by 1492 pixels high.
If the image is projected at 1.85: 1 (the normal American 35mm theatrical projection format ) the visible image is the equivalent of 2048 pixels by 1107.

Jeremy Hughes
05-15-2007, 12:22 PM
I think it's 2,048 x 1,152... not sure, though.

I thought it was. It's 16:9 at all resolutions right?

TimothyD
05-15-2007, 12:30 PM
See, I told you I was no math whiz, just trying to be helpful:)

I probably shouldn't try to help with math...

Damien Molineaux
05-15-2007, 04:17 PM
First we have to figure out if we're talking RGB from the 35mm sized sensor or from the S16 sized sensor. Although in both cases your RGB is but an interpretation of the information from the single CMOS sensor.

This is how I understand things :

So if using the S16 sized sensor, your RGB is actually downsampled from 2k, with some preservation and some interpretation (debayering) added color information. So in fact your are throwing away information and you could just record 2k and gain some 66% (theoretically) in file size. That is the advantage of RAW.

In the case you're recording RGB from the 35mm sensor, well then you're gaining some space (smaller files) compared to 4k raw, although we have yet to see how Red RGB compresses compared to Red RAW. In this case you gain shallow DOF compared to 2k, or you lose DOF if you would rather have more. There is of course a lot less interpretationn going on since you're downsampling to 1080 from 4k, so you actually have more information & detail than 2k. On the other hand, your RGB images are less flexible in post, as far as color goes.

I hope I'm understanding things correctly and that its clear.

Cheers,
Damien

Martin Drew
05-15-2007, 05:20 PM
Another little thing to throw in the mix is that you also have to consider the effect of the optical low pass filter, which will tend to narrow the resolution gap between 2K aquired bayer and 1080p 3 chip still further.

M

Michael Hastings
05-15-2007, 08:26 PM
Adding the channels to find the resolution is flawed, but people make this mistake all the time when discussing 3CCD systems. The RGB channels overlay to create an image with resolution equal to a single sensor.


Most 3CCD's use pixel offset (the pixels of one chip line up between the pixels of the other chip) to increase resolution - it's not doubled, I think it is a square root thing so about 1.4 times the resolution of the single sensor.

Rob Lohman
05-16-2007, 05:48 AM
RED's 2K is 2048 x 1152 indeed.