Specifications
8
A Digital Video Primer
Like computer monitors, televisions also display video using red, green, and blue phosphors.
However, television signals are not transmitted or stored in RGB. Why not? When television was
rst invented, the system was optimized to work in only black and white. e term black-and-
white is actually something of a misnomer because what you really see are the shades of gray
between black and white. With black-and-white television, the only information being transmit-
ted is brightness or luminance.
When color television was being developed, it was imperative that the new system be compatible
with the black and white system, so that millions of people didn’t have to throw out the sets they
already owned. Instead of transmitting RGB, the component signal is converted to something
called YUV. e Y component is the same old luminance signal that is used by black-and-white
televisions, and the U and V components contain the color information or chrominance. e
two color components determine the hue of a pixel, while the luminance component determines
its brightness. With a YUV signal, a color television can reproduce a color image, and a black-and-
white television can simply ignore the U and V components and display a black-and-white image.
YUV is typically associated with analog video, where YCrCb, is used in the digital realm.
Color sampling
When working with RGB images, the same number of bits is used to store the three color
components. When working with YCrCb video, on the other hand, a phenomenon of human
perception is used to reduce the amount of data required for each pixel. Because the eye is much
more sensitive to changes in the luminance of an image than to its chrominance, broadcast-quality
video uses only half as much color information as it does luminance information. Using less
color information helps save bandwidth for transmission, as well as storage space.
In technical terms, the NTSC broadcast specications call for video to provide 8-bit samples at
13.5 MHz with a 4:2:2 sampling ratio. What does all this mean?
• 13.5 million times per second an 8-bit sample of the black-and-white or luminance (Y) compo-
nent is taken.
• 4:2:2 is the ratio between the luminance (Y), and the Cr and Cb color samples. It means that
for every four samples of the luminance (Y) component, two samples of each of the two color
components (Cr and Cb) are taken—360 samples per scan line.
YCrCb can be reduced even further to what is known as 4:1:1 color, in which for every four sam-
ples of the luminance (Y) component, one sample of each of the two color components (Cr and
Cb) is taken—180 samples per scan line. 4:1:1 color provides adequate quality for most consumer
or prosumer (nonbroadcast) needs. e reduced information in 4:1:1 color is not a problem in
most usages, but it can cause issues such as visual artifacts around composited images.
Figure 3 shows what happens when each pixel is sampled from right to le across each horizontal line.
As you can see, in 4:4:4 color sampling, each pixel contains a Y, Cr, and Cb sample. With 4:2:2,
each group of four pixels contains four Y samples, two Cr samples, and two Cb samples. With
4:1:1, each group of four pixels contains four Y samples, one Cr sample, and one Cb sample—180.
You may also encounter 4:2:0 color. is notation does not mean that the second chrominance
(Cb) component is not sampled. In 4:2:0 color, the chrominance resolution is half the lumi-
nance resolution in the horizontal domain (like 4:2:2 color), but is also half the resolution in the
vertical domain. e original 4:2:0 color space is only used for progressively scanned images,
because reduced vertical resolution means that every other line has no chrominance component.
If 4:2:0 were used for interlaced video, then all the color would be removed from the second eld.
Video codecs that use 4:2:0 (MPEG-2 and Microso’s VC1) get around this limitation by using a
modied 4:2:0 scheme, in which the locations of the chrominance pixels are shied so that color
information is evenly divided between elds.
Figure 3: Color sampling
How the Y, U, and V color components are sampled
to convert from 4:4:4 to 4:2:2, 4:1:1, and 4:2:0. Note
that 4:2:0 sampling converts odd and even lines
dierently.