Chroma Subsampling 4:4:4, 4:2:2 and 4:2:0

by Joost | Last Updated: June 27, 2022
I love creating free content full of tips for my readers, you. I don't accept paid sponsorships, my opinion is my own, but if you find my recommendations helpful and you end up buying something you like through one of my links, I could earn a commission at no extra cost to you.

You’ve probably seen the 4:4:4, 4:2:2 and 4:2:0 numbers and other variations, higher is better right?

To understand the importance of these designations, you need to know what these numbers mean and how they affect video. In this article we limit ourselves to the 4:4:4, 4:2:2 and 4:2:0 chroma subsampling algorithms.

Chroma Subsampling 4:4:4, 4:2:2 and 4:2:0

Luma and Chroma

A digital image is made up of pixels. Each pixel has a brightness and a color. Luma stands for clarity and Chroma stands for color. Each pixel has its own Luminance value.

Subsampling is used in Chrominance to use the amount of data in an image sparingly.

You take the Chroma of one pixel to calculate the value of neighboring pixels. A grid is often used for this that starts at 4 reference points.

Luma and Chroma

Ratio formula of Chroma subsampling

The chroma subsampling is shown in the following ratio formula: J:a:b.

Playmobil petting zoo stop motion
Playmobil petting zoo stop motion

J= the total number of pixels in the width of our reference block pattern
a= the number of chroma samples  in the first (top) row
b= the number of chroma samples  in the second (bottom) row

See the image below for a 4:4:4 chroma subsampling

Ratio formula of Chroma subsampling

4:4:4

In this matrix, each pixel has its own Chroma information. The codec does not need to estimate what the Chroma value should be because it is recorded in every single pixel.

This gives the best image, but is reserved for the cameras in the very highest segment.

4:4:4

4:2:2

The first row only gets half of this information and has to calculate the rest. The second row also gets half and has to calculate the rest.

Because the codecs can make very good estimates, you will see almost no difference with a 4:4:4 image. A popular example is the ProRes 422.

4:2:2

4:2:0

The first row of pixels still gets half of the Chroma data, which is enough. But the second row has absolutely no information of its own, everything has to be calculated based on surrounding pixels and luminance information.

As long as there is little contrast and sharp lines in the image, this is not a problem, but if you are going to edit the image in post-production, you can run into problems.

4:2:0

If Chroma information has disappeared from the image, you will never get it back. In color grading, pixels have to “estimate” so much that pixels are created with wrong Chroma values, or block patterns with similar colors that do not correspond to reality.

With a Chroma key it becomes very difficult to keep edges tight, let alone smoke and hair, the data is missing to recognize the colors correctly.

A 4:4:4 grid isn’t always essential, but if you want to edit the image later, it helps to have as much Chroma information as possible.

Work with the highest subsampling values ​​for as long as possible, and only convert to a lower subsampling value before final publication, for example online.