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

Why throw colour away?

Human vision resolves fine detail mainly through brightness. We notice a blurred edge instantly, but a slightly blurred patch of colour much less. Video systems exploit this by keeping full-resolution luma and reducing the resolution of the chroma channels.

Reading the notation

The notation J:a:b describes a block of pixels J wide and two rows tall:

• J — the width of the reference block, almost always 4.
• a — how many chroma samples are stored in the first row.
• b — how many new chroma samples are stored in the second row.

The three common schemes

4:4:4 keeps a chroma sample for every pixel. Nothing is discarded. It is used for graphics, screen content and high-end finishing.

4:2:2 halves chroma horizontally. Each pair of pixels shares colour, but every row has its own. It is common in professional acquisition and broadcast.

4:2:0 halves chroma both horizontally and vertically, so one colour value covers a 2×2 block. Almost all consumer video, streaming and discs use it. At 8 bits, it needs 1.5 bytes per pixel instead of 3.

When you can see the difference

With natural footage, 4:2:0 is usually indistinguishable from 4:4:4 at normal viewing distance. The difference appears with:

• Sharp coloured text and user-interface elements, which develop soft or bleeding edges.
• Chroma keying (green screen), where the key edge becomes blocky.
• Heavy colour grading, which can reveal chroma blocks in saturated areas.

Practical advice

Record at the highest chroma resolution your workflow can handle if you plan to key or grade heavily. Deliver in 4:2:0 for distribution, where it is expected and efficient. And remember that converting from 4:2:0 up to 4:4:4 does not restore lost detail; it only interpolates it.