Image Processing Reference
The second digit (which is often the value “2”) means that during the same sampling
period, only 2 samples are taken for each of the color-difference signals in the horizontal
The third digit (which is also often the value “2”) describes the sampling in the ver-
tical axis but this sub-sampling involves some additional computation to halve the amount
There is sometimes a fourth digit that represents an alpha channel-blending value.
This becomes important for the more advanced codecs described by MPEG-4 and must be
sampled at the same rate as the luma signal or else some unpleasant fringing effects will
appear around the edges of the alpha-blended video object.
Table 5-8 lists some related formats that are relevant: In the table below, the first sam-
pling pattern is called 4:4:4. The full range of values is preserved and this is essentially an
uncompressed pixel map. Luma and chroma are both sampled at full resolution.
Some data reduction is achieved with 4:2:2 sampling, which reduces the amount of
chroma information by only sampling half the pixels. The other pixels are ignored. Note
that this is not an average value of two samples; it is selecting one of two samples.
A further reduction to a quarter of the chroma information is achieved with the 4:1:1
sampling pattern. This ignores 3 out of 4 color values and only records the fourth.
Table 5-8 Luma and Chroma Sample Sizes
Luma and chroma sampled
at full resolution
Luma sampled at 4 times
the sub-carrier frequency.
Chroma sampled at half
the resolution on the
horizontal axis and
down-sampled in the
Double-sized aperture cell
to allow higher
compression ratio for
chroma. 8:2:2 is a more
of the sampling.
Only one chroma sample
for the whole block of