Colour Subsampling – What is 4:2:2 for flip sake?

Colour Subsampling

With the release of the new Canon C100 announced recently everyone will be wondering why it is half the price as its older brother, what exactly do we lose for £5000 less?

Canon C100 Side 02

One very important feature is the codec the C100 captures in.. AVCHD at 24Mbit 4:2:0.. and its this 4:2:0 we will be looking at today to try and understand what it means and if we notice it after its been encoded for vimeo 720p.

We’ve all seen the numbers 4:2:0, 4:4:4, 3:1:1, 3:1.5:1.5, 4:1:1 and 4:2:2 but what the heck do they all mean and will it affect the films and videos we produce?

Most of todays cameras take the light coming through the lens and convert it to 3 sets of numbers, one Red, one Green and one Blue.. this is called RGB, the problem with this data is its MASIVE in size so the camera uses a technique called ‘colour sampling’ or if you cant spell correctly (American) ‘color sampling’ this splits the colours into a new set of numbers called YCbCr or YUV to reduce the size of each file, this makes it easier to stream over the internet, broadcast, record onto CF cards or back in the day.. video tapes (ask an adult)

YUV breaks down the signal into
Y = Luminance (brightness of colours)
Cb = Chroma blue (amount of blue saturation)
Cr = Chroma red (amount of red saturation)

Colour sampling works in a similar way to the human eye which uses rods and cones to sense light, the rods sense brightness, light and dark and shades of grey in-between while the cones see colour but they are fewer in number.  We find it easier to see changes in brightness than change in colour.  Back in the day video engineers used this system on the basis that “we’re not really going to notice anyway”

Where does the 4:2:2 and and 4:4:4 come in?

Have a look at the diagram below and imagine it is part of a larger image, just 4×4 matrix of pixels that are used to make up a larger image:

In the example every box has a Y value (the black square), a Cb value (blue) and a Cr value (red), so in shorthand we would call this 4:4:4 over a set of 4 pixels wide.

Not a lot of cameras shoot in 4:4:4, its normally called RAW and takes up a lot of space on a card or drive. Its also not really noticeable to the human eye when colours are taken out or added.

Most cameras shoot on 4:2:2, this is what it would look like as a diagram similar to before:

As you can see 1/2 of the pixels are missing important colour data but this isn’t incredibly noticeable when you watch it on a HD screen

Basically the higher the number the better the colour data, RED Epic RAW records the RGB data straight from the sensor which is why all the best directors are using them.. It’s as close to film as they can get at the moment to film but everything will change in the next year or so.. it always does.

There are a number of plugins you can get to bring colour data back into your footage and minimise the impact of low colour sampling, if you own a PC search for a ‘chroma blurring filter’ to smooth out the blocky edges, and on a Mac there is a plug-in called ‘Nattress’s G-Chroma Smoother’ which can be found here