7.5.2. Color Subsampling
Digital video uses the Y, Cr, Cb components in the YUV color model to represent
the color of each pixel. Y is a measure of the brightness or luminance of the color.
By convention, Y is derived directly from the G-component of an RGB representation.
Cr is the red chrominance, which is a measure of the difference between the
intensity of the red color component and the green color component. Cb is the blue
chrominance, which is a measure of the difference between the intensity of the blue
color component and the green color component. Since the Cr and Cb components of a
YCrCb color representation represent differences between the R and G components and
the B and G components of an RGB representation, respectively, Cr and Cb can be
The YCrCb system of color representation is used in digital video because the human
eye is more sensitive to the intensity (brightness) than it is to color (hue) for a
group of color pixels. Digital video systems take advantage of this phenomenon to
reduce storage of information by assigning more bits to storing the Y component
than to storing the Cr and Cb components of the color representation. If we encode
just the luminance and discard the two chrominance values entirely, we have the
black and white version of the image. MPEG encoding does not have to be accurate
with chrominance for every single pixel. Instead of storing the chrominance values
for every pixel, it encodes the average chrominance values for a group of four
pixels (usually a square of two-by-two pixels from the original picture) while
retaining the full luminance values for every pixel. This method is called color
After converting from the RGB color representation to YCrCb representation, usually
8 bits (representing 256 shades or intensity levels) are used to represent each
color component. Therefore, each pixel of a video frame is represented with 24 bits
of YCrCb data, just as it is represented with 24 bits of data in RGB format. Without
further processing, this is known as a 4:4:4 sampling. This format is equivalent to
the RGB format since there is no saving in storage (no compression).
In each group of two pixels, the 4:2:2 sampling format discards the Cr and Cb
components of the second pixel. The Cr and Cb components of the first pixel are
used instead. In other words, the Cb and Cr components are horizontally subsampled
by a factor of two (each Cb and Cr sample corresponds to two Y components). This
form of compression results in reduction of storage requirements at the expense
of loss of color information.