Chroma Subsampling

4:4:4 vs 4:2:2 vs 4:2:0

Chroma Subsampling is closely entwined with Color Encoding Systems, they do however have completely different functions but one cannot live without the other.
But what is Chroma Subsampling exactly and how does it correspond with Color Encoding Systems?

Before we begin that journey, it is important to understand just how video works. Most of us just see it as a video signal being transferred to a playback device, but as a matter of fact, it’s not that easy.

The best known Colour Encoding System is probably RGB, you take a colour and convert it into a red, green and blue number and when you recombine those numbers, you recreate the colour. How does that compare to let’s say YUV? It doesn’t. YUV is a Colour Encoding System that takes in account human perception.
As you might know, we are unable to see all the colours there are around us, so why would we spend bandwidth to transport information we cannot see?

That’s why they created a system that works better for us. And that’s where YUV comes in the picture. YUV defines one luma (Y or luminance information) and two chrominance (UV or colour information).
Back in the days when we only had black and white televisions, we only needed the luma. However when we add colour to the picture, we also need chrominance which is represented as two colour difference components, with U being B’ – Y’ (Blue – Luma) and V being R’ – Y’ (Red – Luma).
The issue we have here is that in the YUV format, luma only receives 1/3rd of the available bandwidth whilst being more important than chrominance.
Look at it this way, black and white is responsible for the contrast, shadows,… if you look at a black and white image, it does not appear less sharp than a colour image, if however we look at a colour image without the blacks, it will look very different. Hence the importance.



Left we see the image as we want it and right we see the luma and chrominance separated from each other.



And that’s where Chroma Subsampling comes in, Chroma Subsampling allows us to compress the chrominance so that we have up to 50% more space for luma using the same bandwidth.

But how does it work?

The first number we see in every one of the options is a 4, this refers to the size of the sample, the following numbers refer to the Chroma Subsampling. They are both relative to the first number and give us the horizontal and vertical sampling.
If you use 4:4:4, there is no compression, it is not subsampled. However when you use 4:2:2, it contains half the chroma compared to 4:4:4 and if you use 4:2:0 it only has a quarter of the chroma compared to 4:4:4.

If we look closely at the image above, we see Chroma Subsampling in action.
As mentioned before, 4:4:4 has no subsampling.

If we take a look at 4:2:2, you will notice that the luma will not be touched and that the horizontal sampling will be halved while maintaining the complete vertical sampling.
With 4:2:0 on the other hand, ┬áboth horizontal and vertical sampling will be halved, this means that it will sample the colours out of the first row’s pixels and ignore the second row completely.

In movies and TV-shows, subsampling is always done with 4:2:0 and does not have a visual impact.


Leave a Reply

Your email address will not be published. Required fields are marked *