Comments on: Color Subsampling, or What is 4:4:4 or 4:2:2??

By: Karl Soule

Karl Soule — Mon, 21 Nov 2011 23:52:54 +0000

Jean, thanks for the clarification. You are correct. When writing this article, I was targeting the average editor, trying to simplify a complicated topic.

By: Jean

Jean — Mon, 21 Nov 2011 23:47:04 +0000

First excuse my bad english. I speak french.
Your explanations are not entirely accurate. In 4:2:2 for ex. it is not thrue that the color information is discaded for each even pixel. In fact, the color information is subsampled by 2. So, if the resolution of Y is 720 x 576 (Pal), the resolution for U and V is subsampled to 360 x 288. The firts and second Y pixels receive the U an V values the first pixel from the subsampled color information. Thus, the first twoo Y pixels receive the same color value (the average of their actual colors), and so on. If you convert now to 4:4:4 and back to 4:2:2, there is absolutly no modification in the color information. In an uncompressed YUV (YUY2) file, the values for the first pixel is coded YUYV and so on, with the U and V values valid for the 2 Y.

By: Adam

Adam — Thu, 28 Apr 2011 22:26:46 +0000

I work with some video experts but I rarely get the opportunity to question them extensively so I dig online and read books. I must say nothing puts this stuff in plain english as well as what I am finding here. Thank you!

So my comment / question is this: It makes sense that converting from 4:2:0 to 4:4:4 then back to 4:2:0 to many times would mess up the picture because the conversion function does not know where the original color subtractions were performed… Or does it? If the resolution remains the same (width and height), would it not make sense that the color subtraction is always in the same positions? Go from your first illustration to your second and back. Nothing is lost. So as long as there is a standard way to choose the pixels that will lose color information then conversion routines would be “safe” would they not?

By: Karl Soule

Karl Soule — Tue, 02 Nov 2010 16:08:10 +0000

Cineform is still an excellent codec, and I still use it when I need to work both inside and outside of the Adobe Production Premium suite. I also LOVE the color controls and look-up tables (LUTs) available in First Light, and those color changes show up immediately in Premiere Pro.

Since Premiere Pro can do the upsampling to 4:4:4, 32bpc right in the timeline, it’s not a crucial step, but it can still add value to your pipeline.

By: Éric ramahatra

Éric ramahatra — Sun, 31 Oct 2010 18:51:01 +0000

Hi Karl

This was the kind of information I was looking for as I’m tired explaining why it’s not necessary to transcode to 4:4:4 whatever your footage codec is.

In fact grading is codec-independent as we grade images not codecs or file-formats.

It’s like adding effects on a old K7 output or burn the music on a CD first…your effects board doesn’t bother the source material, K7, CD, SACD…it just takes the sounds as is regardless it is 16khz/8bjts, 44khz/16bits or 96khz/24bits…

By: Richard Allen Crook

Richard Allen Crook — Thu, 28 Oct 2010 12:56:52 +0000

I used to use Premiere cs4, and I would convert my 5d footage to Cineform Avis for ease in editing. Now that I have cs5, Converting does not speed up the editing anymore so I asked the techs at Cineform if there was any reason to convert. They told me that when color grading it’s best to start with Cineform because the 4:2:2 is better to color grade with in Premiere. Is this not the case? I would love to remove this step from my workflow.