Anyone who wants to have a say about video formats must know a lot of abbreviations and terminology. And that knowledge can be important if you want to get the best out of your television. For example, in the menus of (Ultra HD) Blu-ray players you will sometimes find cryptic settings for the video format: 4: 4: 4/4: 2: 2/4: 2: 0. This strange code specifies the chroma subsampling of the video signal. In this background article Chroma subsampling explained in details.
Chroma subsampling explained: Origin
Chroma subsampling is one of the many compression techniques used in both video and photo. Some of the information is discarded to make the file smaller. This is especially important for video because the files are very large, so the savings are proportional. A second reason why this is especially important for video is that the bandwidth of transmission channels (such as your internet connection) is limited. You couldn’t possibly stream an uncompressed video file. Even for your HDMI cable, an uncompressed video file is just too big to get every frame on your screen in time. Compression is therefore indispensable.
The basis of this technique is very simple: our eye is less sensitive to spatial variation in color than to spatial variation in luminance (brightness). To take advantage of this, images are not saved in RGB encoding (where each pixel is assigned a value for red, green and blue), but in YCbCr. So each pixel still has three values. Y is the value for the luminance, Cb and Cr are the Chroma (color) values. We spare you the math of the conversion between RGB and YCbCr.
When we visually split a photo according to this model, it immediately becomes clear that the luminance channel is much more important.
The amount of detail in the luminance channel (black and white image) is sufficient to have a clear impression of the image. The chroma channels are much less clear.
Chroma subsampling explained: How do you compress color data?
Now that the Chroma information is available separately, it is possible to compress that information. The method used is sub-sampling, in other words less samples are taken than are actually needed. To what extent there is too little sampling depends on how much you want to compress and how much color information you are willing to lose. For example, the number of chroma samples can be halved horizontally or vertically, or both. The notation that indicates this is, to say the least, not very intuitive
The general form is J: a: b, and this describes how many chroma samples are taken per block of Jx2 pixels. In the first row of that block, we sample a pixels for color. In the next row of this block, we sample b pixels for color. As a reminder, luminance is always sampled at every pixel.
A few examples make this much clearer.
4: 4: 4
This is uncompressed video. Four chroma samples in the first row and four in the second row.
4: 4: 0
Four chroma samples in the first row, and none in the second row. This means that the chroma samples from the first row are also used for the second row. In other words, we halve the vertical resolution, but keep the full horizontal resolution.
4: 2: 2
Two chroma samples in the first row, and two in the second row. Each chroma sample is used for two columns. This scheme cuts the horizontal resolution in half, but maintains full vertical resolution.
4: 2: 0
Two chroma samples in the first row, and none in the second row. Each chroma sample is used for two columns and for the row below. This scheme halves the horizontal and vertical resolution.
There are other schemes, some of which even deviate from the conventions described above, but unless you are a professional with video, it is sufficient to know 4: 4: 4, 4: 2: 2, and 4: 2: 0. It is these schemes that you regularly encounter in cameras, and the settings of players.
Chroma subsampling: Image artifacts
What is the effect of chroma subsampling on the image? The impact is always most visible with a bold transition from one color to another. Since only one sample is taken at the border, the two colors mix. In the case of 4: 2: 0 where you halve the resolution both vertically and horizontally, the color expands and dilutes even more. In the example below it seems very painful, but keep in mind that these are huge pixels.
A second example still shows large schematic pixels. Move a little further away from the monitor, or focus your eyes further and the difference suddenly becomes a lot less dramatic.
he images below show the possible effect on more realistic content, still enlarged so that you can see the effect.
It may be clear that chroma subsampling does have an effect on image quality. Particularly in very fine and pronounced color patterns, it is possible that some colors will fade. Also, be aware that the method for bringing a subsampled signal back to uncompressed 4: 4: 4 video is not fixed. Thus, different devices can produce different results.
Almost all the content we watch every day uses 4: 2: 0 chroma subsampling: DVD, Blu-ray and even Ultra HD Blu-ray, as well as TV broadcasts, YouTube, Netflix and so on. Some cameras can record in 4: 2: 2. This is an advantage, especially for those who still want to do post-processing. Even then, your final video will almost always be in 4: 2: 0 format.
However, HDMI only accepted 4: 4: 4 and 4: 2: 2 up to and including version 1.4. That means that each player had to do a conversion internally from 4: 2: 0 to 4: 2: 2. The television then takes the further step to 4: 4: 4. To avoid the video having to go through two conversion processes, some players allowed to convert the video stream from 4: 2: 0 to 4: 4: 4 immediately.
Since the introduction of HDMI 2.0, it is also possible to forward 4: 2: 0. So there is no need to do a conversion in the player anymore. Yet many players still offer this. The question is then which of the two devices will perform the conversion better, the player or the television. In that case you can try to see a difference yourself. In many cases, we still recommend leaving these types of settings on ‘Auto’, so you can be sure.
An interesting fact: if you choose 4: 4: 4 on the player, you send a lot of data through the HDMI cable. If in that case (and that is probably only with Ultra HD content) you see asterisks in the image (individual pixels that flicker) or if the image is occasionally lost, this means that your cable cannot handle that amount of data. By switching the player back to 4: 2: 2 or 4: 2: 0, you lower the data rate and everything may be fine again. If that is not the case, then you need a different HDMI cable .