is RED 4:4:4 or 4:2:2?

[cinemano]

Sorry for the silly question.. new with this ..

[Gunleik Groven]

You should read Graemes "Why Bayer" thread...

[Radoslav Karapetkov]

Definitely, it's a must-read:

http://www.reduser.net/forum/showthr...ight=why+bayer

[luis bustamante]

From the Red FAQ:

"How does the Red One differ from a traditional HD cam?

HD cams come in many flavors, recording anywhere from 720P resolution to as high as 1080P, using anything from a standard definition sensor with interpolation to a true 1920 x 1080 pixel sensor. In the very highest quality HD cams, the RGB signal is not subsampled, offering the best HD quality at 4:4:4 (HDCAM SR). Typically, the chroma is subsampled at 4:2:2 RGB (DVCPRO HD) and 4:2:0 (HDV). Effectively, where HD leaves off, the RED One begins, offering 2K, 3K and 4K recording options. RED uses a sensor measured at 4520 x 2540 pixels, more than 4 times the number of pixels of the very best HD cam and records the signal as a RAW input, similar to a DSLR, with no color subsampling. All information travels in a single channel, as opposed to three separate RGB channels. Where a video camera requires the setting of white balance, color and gain before recording, these are all adjustments that can be made to a RAW recording after the fact, thus making the process of capture all the easier. The greatest benefit to the cinematographer is the fact that the sensor on a RED One is Super 35 sized at 24.4mm x 13.7mm, offering 35mm Depth of Field."

[David Mullen ASC]

The point is that you can create either 4:4:4 / RGB or 4:2:2 YCbCr from the RAW images that the camera generates.

[Graeme Nattress]

There are many answers to this question.

1) RED is RAW, hence such designations of chroma sampling do not apply

2) Chroma subsampling notation is for when, as part of the compression of the image, chroma data is reduced in comparison to the resolution of the luma. RED does not do this in it's compression.

3) Some people want to apply such notation to sensors. It never has and never will be a sensor notation.

4) Some people want to apply such a notation to measured chroma / luma, and again, it's not the correct notation for that.

RED uses a Bayer Pattern Colour Filter Array sensor - see my post on that subject. The reconstruction of the RAW to RGB produces a 4:4:4 RGB image. Because of the optical low pass filtering and because we use a Bayer pattern with all it's advantages in allowing us to make small files of high quality and use PL Mount lenses etc. we cannot extract out from the RAW the full resolution of the pixel count, but we can get >78% out, making a superb image in the process, at a higher resolution than any buyable and usable motion imaging camera.

Graeme

[Andrew M.]

I have one question when it comes to RGB

Once REDCINE converts REDRAW to 16 bit TIFF is this TIFF color space Apple RGB or Adobe RGB (1998) or something else?

In AE when I start project it is asking me, what color space I want to work in. I am picking Apple RGB.

[Graeme Nattress]

Well, normally the colour space is CameraRGB. I'd probably select the target colour space in AE, which would be REC709.

Graeme

[Andrew M.]

Yes, target 709 it is what I am selecting, but source RGB (AE has Source and Target setting)?

I have seen CameraRGB in REDCINE and I was wondering what is the closest color space to it?

[M Most]

2) Chroma subsampling notation is for when, as part of the compression of the image, chroma data is reduced in comparison to the resolution of the luma. RED does not do this in it's compression.

3) Some people want to apply such notation to sensors. It never has and never will be a sensor notation.

4) Some people want to apply such a notation to measured chroma / luma, and again, it's not the correct notation for that.

It's also not the correct notation for a description of compression. Data compression and subsampling are two completely different things. Subsampling may reduce the amount of data that is to be recorded, but it has absolutely nothing to do with data compression. It is not correct to say that the "effect" of subsampling is part of compression simply because it allows for a smaller amount of data in the data stream, because data compression is a specific, separate step applied after the information is sampled.

If we're going for correctness, let's get it completely correct.