Final Cut Pro HD | H•O•T Hands-On Training [Electronic resources] نسخه متنی

Why Color Space Drives Me Nuts

On the surface, computers and video should work well together. They both show movement, audio, and color images, so you'd think they'd be compatible.

You'd think so, but you'd be wrong.

As you first learned in Chapter 9, there are five major areas where computers and video diverge significantly:

Scanning:
Computers display progressive images, showing every horizontal line consecutively. Video uses interlaced images, showing first all the even-numbered lines, then all the odd-numbered lines. This has a significant impact when working with movement inside an image.

White levels:
Computers create images on a digital scale, ranging from 0 to 255. Digital video creates images also on a digital scale, but broadcast video uses an analog scale, which converts to digital values of 16 to 235. This means it is very easy to create graphics on the computer that are too white, or too saturated, for video.

Pixels:
Computers use square pixels. Video uses rectangular pixels. Great-looking images on the computer look tall and thin on video.

Color space:
Computers use RGB. Video uses YUV. YUV does not display the same colors as RGB. Video is especially weak displaying saturated shades of yellow and blue.

Color space compression:
Computers display all pixels individually at full value. Video compresses color information to reduce file size and make transmission easier.

In this example, imagine a block of pixels, showing a blonde actor in front of a green screen. Each of these pixels is defined by three numbers: Y defines the luminance or gray scale value; U and V define the color.

When each of these four pixels is described using its own Y, U, and V values, that image is defined as using 4:4:4 color space; four pixels using four Y values, four U values, and four V values. This is the normal state for your computer, but it is never used in video; the file sizes and transmission requirements are too big. (4:4:4 DV video, if it existed, would require a data rate of slightly more than 31.1 MB per second for playback.)

The human eye is very sensitive to changes in luminance, that is, to changes in brightness. So, a long time ago, video engineers decided not to mess with the luminance value of a pixel, because image quality would suffer too much. However, they still needed to reduce the file sizes of color images, so they started playing games with the color information.

What they decided to do was to average the color values of two adjacent pixels. Rather than display them as discrete values, they would blend them. Now, that four-pixel block had four discrete Y values, but only two discrete U and V values, because colors were averaged over two pixels. This reduced the data rate to about 22 MB per second.

When it came time to create DV, keeping image quality high was important; however, using small file sizes was even more critical. To do this, engineers again compressed the color space by averaging the color values of four pixels. So, now, each pixel in that group of four pixels had a discrete Y value, but shared one common color value. This is called 4:1:1 color space. This reduced the data rate to about 4 MB per second. And, although DV video uses additional compression techniques to minimize the data rate, this color compression is the one with the most direct impact on effects.

Normally, this color compression is not a big deal. To our eyes, DV video looks pretty good. However, when you start doing color correction, or chroma key, or just messing with colors, you very quickly learn that DV video just doesn't have enough color information in it to create good-looking effects.

So, I'll show you how these tools work. But, if creating high-quality effects is your goal, you'll need to move to a higher-quality color format than DV, such as Betacam SP or DigiBetacam.