8.0

Chroma format
The color value of each image point consists of the color values for the primary colors red, green, and
blue (RGB), and for traditional and technical reasons it is transformed into one brightness value (Y =
0.299*R = 0.587*G + 0.114*B) and two color difference values (U = R - Y, V = G -Y).
The Y value alone produces the black and white picture. These signal components allow brightness and
color information to be handled separately. The first data reduction occurs when single rows comprising a
picture are read. Because the human eye has a lower color resolution than a brightness resolution, the
color components are recorded only for every other point of a row (4:2:2) for each four pixels grouped
(4:1:0), i.e. color signal under-reading.
4:2:2
This corresponds to the established TV standard. One piece of color information is transmitted per row
for two pixels which corresponds to a 2/3 compression of the output data.
4:1:0
This is the color coding used for DVDs and most other consumer video applications. For each 4 pixels
grouped together on two rows, one unit of color information is saved. This corresponds to a output data
compression of 1/2.
Field
A half-image, i.e. two halves which combine to produce a frame (see de-interlacing
).
Frame
A frame is a single image from a video sequence which also called a full image. PAL video, for example,
contains 25 frames per second, NTSC 29.97 frames.
Video recordings, with the exception of computer animations and still frames, don't contain full images.
Instead, they have double numbers of half-images (fields) which are transmitted in an interlaced state.
However, we still refer to frames, since many predecessors of MPEG compression are based on such
frames. Video editing literature usually refers to frames.
GOP
Group of P
ictures: The sequence of I frames and the P and B frames that belong to them.
e.g. I B B P B B P B B I ...
(This GOP has a length of 9, with 2 P frames and 2 B frames)
I frames contain the entire image information of a frame, while P and B have part of the information.
So-called prediction
and movement approximation are methods used for reduction.
The combination P B B is called a subgroup.
I frames must appear in regular intervals in the data stream for image and sound to be synchronized.
Between the I frames only a limited count of P and B frames is allowed. This explains a few things: Since
P and B frames contain only differential information, these differences will be larger with time, since more
and more changes takes place from frame to frame. A large count does not make much sense, since
GOP has a maximum length of 15 (4P, 2B) in PAL and 18 (5P, 2B) in NTSC. (More than 2 B frames
between P frames is not allowed).
In a closed GOP
, B frames of the last subgroup may contain only backward predictions or references to the preceding P
Page 417