Re: understanding the jargon of video

[smthopr]

Bit depth (8,10,14 etc) vs color resolution (4.0.0, 4.2.0, 4.4.4)

Bit depth refers to the number of steps in luminance. 8 bit has 256 levels of Red, Green, and Blue. Or even 256 luminance, and 256 for each color channel (like LAB space).

4.2.2 etc refers to the number of pixels recorded in each channel. Ie. 4k for the luminance channel and 2k for the color channels.

You can have 10 bit 4.2.2 or 8 bit 4.2.2 or even 16bit.

What is most important for color correction is the bit depth. 4.2.2 vs 4.4.4 only effects the resolution of the image and not the bit depth, which is important for grading.

Color resolution is very important when doing green screen composites though. 4.4.4 will pull a more detailed matte.

Hope this helps clear this up:)

[Graeme Nattress]

I wrote this a long time ago now, but it's still a pretty good description of how chroma sampling works: http://www.kenstone.net/fcp_homepage/chroma_investigation_nattress.html (you may want to just ignore the discussion of SD analogue video compared to digital that makes up the rest of the article)

Chroma subsampling is a completely independent factor to bit depth.

[smthopr]

I should add that I'm not sure that shooting 4k 4.2.2 and scaling to 2k will result in a 2k 4.4.4 file. All channels might well be scaled in this operation and result in a 2k 4.2.2 file.

It should be possible to accomplish this, but I don't know that the scaling algorithms are written to do this.

[Graeme Nattress]

Generally speaking, a specialized scaler of video, working in the video domain may keep the data in YCbCr and scale the channels, especially if the output is in the same chroma sampling as the input.

But if we're in an app that works in the RGB domain, the video will first be converted from YCbCr to RGB where the resolution of the three channels are equal (up sampling the lower resolution chroma channels). Maybe some smoothing will occur to help the chroma sub-sampling look better too. That full resolution RGB will be then scaled to the target size and thus yes, you will get the equivalent of 4:4:4 sampling if you're scaling 4k 4:2:0 to 2k.

[NancyP]

Well, thanks for the original article and for explanations. I have videophobia, partly because of all the jargon, partly because I am used to thinking about stills composition, etc, and I am wanting to develop skill in one area before dipping a toe into a much more complicated area.