Is it 4K pixels or 4K samples?

[redhead]

Hello, this is my first post on this forum. I am a little unclear on what exactly the 4K refers to. Is it the raw samples, or is the full RGB value per pixel?

In other words, is it 2K green + 2K red in one line, then 2K blue + 2K green in the next line, or is it after the raw format is converted, so there are 4K of full-color pixels per line?

[David Mullen ASC]

"4K" refers to the horizontal number of pixels. A 4K frame is typically 4096 pixels across -- vertical height (and thus total pixels) depends on the aspect ratio.

The camera has a single 4K Bayer-filtered sensor and you record the RAW output. From that, the file has to be de-Bayered to create 4K RGB files (4K for each color record). There is some controversy surrouding the issue of whether 4K Bayer ends up truly 4K in resolution -- it mainly depends on the quality of the de-Bayering algorithm. But the end result, regardless of whether you think there was a resolution hit, are true 4K RGB files after de-Bayering. Most people think the results look comparably sharp to 4K RGB scans of 35mm, which is sort of the standard. See:

http://en.wikipedia.org/wiki/Bayer_filter

I suppose there is some irony that the inventor of Bayer filtering worked for Kodak...

[Graeme Nattress]

Good answer David!

Also, the major resolution reduction is not the bayer pattern but the optical low pass filter, and this is necessary for all cameras, otherwise aliassing results which looks like awful computer graphicy stairsteps on straight sharp edges.

I like to think of the bayer pattern are particularly clever image compression....

Graeme

[redhead]

From that, the file has to be de-Bayered to create 4K RGB files (4K for each color record).

Thank you, David. That is what I was hoping to hear. I am not worried about the controversy. Just wanted to clarify whether the final result is 4K RGB, and you have clarified that.

Thanks.

[Brook Willard]

This FAQ is a good place to start: http://www.reduser.net/forum/showthread.php?t=1487

[redhead]

Oh, the FAQ was the first place I checked. The thing I was asking, not answered in the FAQ, is whether there are 4K sensors per line or whether the 4K is the final resolution after the RGB resolution.

Perhaps I should restate the question:

To obtain the 4K resolution, does it take a green and blue sample from one line, then a red and blue from another and combine them into a single pixel? In other words, does it take four samples per pixel? In yet other words, are there 8K physical samples per physical line? Four samples per RGB pixel?

Or does it take 2K blues samples and 2K green samples in one line, then 2K red samples and 2K green samples in another and then interpolate? In other words, just one actual physical color sample per pixel?

[Stephen Williams]

Oh, the FAQ was the first place I checked. The thing I was asking, not answered in the FAQ, is whether there are 4K sensors per line or whether the 4K is the final resolution after the RGB resolution.

Perhaps I should restate the question:

To obtain the 4K resolution, does it take a green and blue sample from one line, then a red and blue from another and combine them into a single pixel? In other words, does it take four samples per pixel? In yet other words, are there 8K physical samples per physical line? Four samples per RGB pixel?

Or does it take 2K blues samples and 2K green samples in one line, then 2K red samples and 2K green samples in another and then interpolate? In other words, just one actual physical color sample per pixel?

Hi,

2K Green + 1K Blue + 1K Red = 4K

Stephen

[redhead]

Hi,

2K Green + 1K Blue + 1K Red = 4K

Stephen

Thanks!

[Stuart English]

"2K Green + 1K Blue + 1K Red = 4K"

I'm sure that it wasn't meant to imply anything eles, but to be clear, that is just a simple math equation that describes the number of RAW data sampling points on each and every line of the sensor.

This is not even close to the complexity of the math that realizes the final 4K RGB samples.

[Graeme Nattress]

Yup, the complexity of the code that turns that into an RGB image is more immense than that! Also, each colour does pick up the other colours also to an extent, so each sensor is biassed towards a colour, either red green or blue but also records the rest too, at a lower level.

Graeme