Sunday, May 10, 2009

Uncompressed 4:2:2 Sucks

There's this "fad" out there in the "professional" video world to call certain types of video "Uncompressed 4:2:2" which while it could arguably be true, is completely misleading so for the sake of everyone involved (especially the people who are so scared of the word "compressed" and what it does to their precious footage) I'm going to explain why using that term makes you an asshole.

Compression, digital video swear word that it is, means that the computer figures out nifty ways to not keep the redundant data. To explain, we'll count ducks. There are five ducks in front of you. You could write down on a piece of paper "one and one and one and one and one" or you could write "five". You've used some of your knowledge and intellect to "compress" how those ducks are counted. Most compression formats are "lossy" which is to say they throw out visual data to varying degrees to save space ("approximately five ducks"), often significant amounts of space. This is where people get their panties in a bunch, as they don't want a reduction in visual quality.

(lossy) Compression = loss of visual quality = bad, or so the saying goes.

"4:2:2" is a form of pixel subsampling. There are three channels in a video signal, usually referred to as YIQ or YUV (there's a difference between those two, but it's heady). The first channel (the Y) is the luminance of the image. The second two add the color. In a 4:2:2 subsampled image, half the color information is thrown out. This is done for bandwidth reasons just like compression is. It also, just like compression, makes for a poorer final quality image.

So if you are reassuring a client that your "Uncompressed 4:2:2" or whatever is uncompressed you are an asshole.

The caveat to all this nonsense is that it's very hard to handle true uncompressed, non-subsampled imagery, and in many cases it genuinely does not matter. If you're going to edit something and do a light color grade, you'll be fine and no one will be the wiser. If you plan on working with the footage in a visual effects context you will likely run into some issues.

No comments: