User manual

130 MPEG glossary
www.magix.com
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 Pictures: 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 (view page 133)
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 frame, but no
references to the following I frame, since it belongs to the next GOP.