What is Chroma Subsampling?

Today’s subject is about chroma subsampling, which is a compression technique used for image and video files. You might have heard about it with terms like 4:4:4, 4:2:2 or 4:2:0, which are the three common schemes used with chroma subsampling. But what is it?

History

Back in the 1950s, engineers were given the task to find ways to compress video for the widespread use of color television without exceeding the transmission bandwidth. So, just like smart engineers do, after some research, they came up with a clever solution that is still used today. The human eye sees differences in black and white better than in color, so they figured that the information about the brightness of an image is more important than the color information. They created a system where the information for luma, which is brightness, is saved for every pixel of an image. But the chroma information which is color is shared across two pixels or more. That was a very clever approach, and they called it chroma subsampling.

What is 4:4:4

Let’s take a look at the before mentioned schemes. Chroma subsampling divides an image into blocks with a height of two pixels. The width is given by the first number of this scheme. In the case of 4:4:4, that’s four. So each block is four by two pixels.

Now the brightness information gets separated from the color and chroma subsampling is only applied to the color channels. The second number tells you how many chroma samples are taken in the upper line. In our case again a four, which means that every pixel in the first row gets its own color value. The third number tells you how many samples are used in the bottom line. Again, four. So every pixel gets its own color here as well.

That technically means that 4:4:4 isn’t even chroma subsampling, but rather the definition of no chroma subsampling being used here. No information is thrown away.

What is 4:2:2

Next in line is 4:2:2. So we do the process again. First number is the width 4 again. Now the second number is two. So only two samples are taken on the upper line. That way the first two and the last two samples are sharing the same color, even though they might have had different colors before. But remember the luma information is preserved for every of these pixels The third number is a two as well. And you know the rule already. Only two color samples for the bottom line, so pixels need to share color again.

What is 4:2:0

The last common scheme is 4:2:0, and you know the drill already with is for two color samples in the upper line and zero for the bottom line means it’s just copying the color samples from above. So with 4:2:0 you end up with eight different brightness samples, but only two color samples, which saves up a lot of space. But chroma subsampling is a lossy compression, which means you are losing information and therefore detail and quality. But as said earlier, your eyes probably won’t notice this trick, especially with higher resolution formats like 4K where you can’t really tell pixels apart.

What is 4:4:4:4

Back to chroma subsampling. You might have seen schemes having four numbers like 4:4:4:4, which is also one of the recording options in our live video production app Mimo live. The new fourth number is the alpha information, which is used if your video supports transparency, you’ll know when you need to use that, but in most cases, you’re good without an alpha channel.

How to Choose

Anyway, here are some situations where it does make a difference, and you should consider using 4:4:4 footage. In some cases, a line or block pattern can be distorted because when the pixels share the color, the pattern is gone. But this rarely happens because the pattern has to be so fine for the color to be merged. Your computer is usually sending the video signal to your monitor in 4:4:4, because in certain situations, text can look really blocky when subsampled. And so to ensure everything looks as awesome as possible, subsampling isn’t applied here.

If you’re shooting a clip in front of a green screen, shooting 4:4:4 if possible. Because when we’re moving the green pixels, merged pixels at the edges of your subject will affect the end result and make it mushy.

Conclusion

Let’s quickly take a look at what you are saving with chroma subsampling. 4:4:4 is our starting point, and consider it 100% when converting that to 4:2:2, you only need 66.7% as much data. And when using 4:2:0, you are down to 50% of your original data. That’s half of what you started with.

Now you’re one step closer to master your video production workflow and understand things like ProRes 4:2:2 a little better.

Darren

Filmmaking tips & tricks, camera gear reviews, cinematography breakdowns, and other training-based content to help you level up your filmmaking.