As our industry dives headlong into the digital video domain, content producers and video engineers find themselves dealing with color sampling ratios on an ever increasing basis. In a nutshell, a "color sampling ratio" is a precise numeric method of describing the relationship between the luminance and chrominance values in a digital video signal. Currently, there are many digital video formats in use, none of which are "perfect," and all of which include some form of signal compromise.
In order to select the best compromise between the different digital video formats, we need to understand what the "numbers" mean in the color sampling ratio itself. For this discussion, it will be assumed that we are discussing video signals that are similar in resolution and bit depth (future tech tips). In the final analysis, all three of these factors determine the final bandwidth size and overall quality of the image.
"Perfect" video (or at least, video that doesn't include sub-sampling), is represented by the ratio 4:4:4. The three-digit sequence represents the ratio of how the Luma (Brightness), Chroma Red (Cr), and Chroma Blue (Cb) components of the signal are sampled, and how that sampling relates to a pattern of four pixels. Specifically, each digit represents the number of times that value is sampled within the four pixel zone. Thus, given a 4:4:4 ratio, each signal component is sampled four times in a four pixel sequence - and no samples are discarded.
Perfect, and Professional
RGB signals from a PC are generated at 4:4:4. There are several high-end camera systems that output at the same 4:4:4 sample rate. As the ratio implies, no samples are discarded, thus there is no loss of quality. The downside is that a 4:4:4 signal requires a great deal of bandwidth, which makes it impractical for consumer broadcasting and normal recording. To allow for video transmission, and to enable longer record times on a hard disk recorder, the sample rates must be reduced.
The professional standard is 4:2:2. Here, the two chroma components are sampled at one half the rate of the luminance component. As a result, the bandwidth of the overall signal is reduced by one third. Even though the information is compromised, the viewer notices little or no signal degradation as a result of "discarding" samples. In practice, the 4:2:2 sample rate is used in Betacam, D-1 and CCIR 601/SDI products.
Heading Down the Ratio Tree
A further reduction in the sample rate yields the 4:1:1 scheme. Here, the chroma channels provide one fourth the bandwidth of the "perfect" 4:4:4 sample rate. Although this ratio is widely used for ENG (Electronic News Gathering) and for industrial video production, the format has limited professional broadcast acceptance. In practice, the 4:1:1 sample rate is used in DV, DVCPRO and DVCAM systems.
Stepping down the ratio tree, the 4:2:0 sample rate is widely used in MPEG-2, DVD, HDV and JPEG implementations. Because the Cr and Cb components are sub-sampled differently, this scheme has the benefit of low bandwidth for easy transport and storage, but you can also end up with increased artifacts and bleeding colors.
Quality Versus Bandwidth
True, there are even more variants, but the schemes presented above are the most common video sampling rates in use today. As we move down from "perfect video," the content producer has to weigh the compromises in quality versus bandwidth, when deciding which color ratio to deploy in a digital video workflow. In the Rental and Staging world, this workflow ends with the content being displayed on a large projected image or an LED wall.
Clearly (pun intended), the lower the sample rate, the worse the image appears, especially if you're scaling the image to a larger resolution. When scaled up, a full 4:4:4 image appears clean, with little color bleeding between the pixels. Presentation switchers that output scaled video to a display device like a projector, typically output 4:4:4.
As you go down the ratio tree, luminance sampling always stays the same, but what changes is the way that color channels are sampled. As we reduce the sampling, colors start to bleed and smear, artifacts are magnified, and this is one of the reasons why Betacam (at 4:2:2) looks better than DVD (at 4:2:0).
When you scale an image, its basic (native) sample rate remains constant. If the information is not there to begin with, it won't be there as it is scaled. It follows the old axiom, garbage in gives you garbage out. The best processor in the world can not create color information that is not there, it can only clean up the information it has on hand.
Choosing a Format
These compression ratios all work on the principle that the human eye is more sensitive to intensity (luma) than to colors, which is why luminance always remains at "4." The final size of the image and the final display on which it appears will help you determine the best format to use in your workflow.
As a rule, any content that you create which will be sent to a presentation screen, should be created at the highest sample rate possible (given your budget), such as Beta, D-1 or SDI. As mentioned earlier, artifacts generated by the 4:2:2 standard will not be seen by the average viewer. In the larger budget productions, issues such as hard drive space or broadcast bandwidth are not as critical as high image quality.
You can certainly use the 4:1:1 sampling scheme (found in the DV/ DVCAM format), provided that you are aware of the quality/bandwidth compromises. The 4:1:1 rate enables you to use less expensive recorders, media servers and cameras, and you can certainly store more content for your presentation. However, you need to be careful of the amount of scaling that you perform on the native signal. As the image is manipulated for the display, it may start to degrade, becoming blurred and pixilated.
As for 4:2:0, this format is best used for archive recording or web casting. This is the format of choice for many consumer level video products and is also the format used for all MPEG-2 video. Its low sample rate allows for easy "transmission" through the Internet, or extended length playback on DVD or VHS. There are industrial level HD camera systems utilizing this sample rate, for example, the Sony XDCamEX. Combined with a High Resolution Imager (CCD), this lower sample rate system still produces a high quality, visually acceptable video signal.