<s>
In	O
the	O
field	O
of	O
video	O
compression	O
a	O
video	O
frame	O
is	O
compressed	O
using	O
different	O
algorithms	O
with	O
different	O
advantages	O
and	O
disadvantages	O
,	O
centered	O
mainly	O
around	O
amount	O
of	O
data	B-General_Concept
compression	I-General_Concept
.	O
</s>
<s>
B‑frames	O
can	O
use	O
both	O
previous	O
and	O
forward	O
frames	O
for	O
data	O
reference	O
to	O
get	O
the	O
highest	O
amount	O
of	O
data	B-General_Concept
compression	I-General_Concept
.	O
</s>
<s>
Three	O
types	O
of	O
pictures	O
(	O
or	O
frames	O
)	O
are	O
used	O
in	O
video	O
compression	O
:	O
I	O
,	O
P	O
,	O
and	O
B	B-General_Concept
frames	I-General_Concept
.	O
</s>
<s>
An	O
I‑frame	O
(	O
Intra-coded	B-General_Concept
picture	I-General_Concept
)	O
is	O
a	O
complete	O
image	O
,	O
like	O
a	O
JPG	O
or	O
BMP	B-Device
image	I-Device
file	O
.	O
</s>
<s>
P	O
and	O
B	B-General_Concept
frames	I-General_Concept
are	O
also	O
called	O
Inter	B-Algorithm
frames	I-Algorithm
.	O
</s>
<s>
The	O
order	O
in	O
which	O
the	O
I	O
,	O
P	O
and	O
B	B-General_Concept
frames	I-General_Concept
are	O
arranged	O
is	O
called	O
the	O
Group	B-General_Concept
of	I-General_Concept
pictures	I-General_Concept
.	O
</s>
<s>
A	O
frame	O
is	O
a	O
complete	O
image	O
,	O
and	O
a	O
field	O
is	O
the	O
set	O
of	O
odd-numbered	O
or	O
even-numbered	O
scan	B-Device
lines	I-Device
composing	O
a	O
partial	O
image	O
.	O
</s>
<s>
Frames	O
encoded	O
without	O
information	O
from	O
other	O
frames	O
are	O
called	O
I-frames	B-General_Concept
.	O
</s>
<s>
B-frames	B-General_Concept
use	O
prediction	O
from	O
a	O
(	O
possibly	O
weighted	O
)	O
average	O
of	O
two	O
reference	O
frames	O
,	O
one	O
preceding	O
and	O
one	O
succeeding	O
.	O
</s>
<s>
In	O
the	O
H.264/MPEG	B-Application
-4	I-Application
AVC	I-Application
standard	O
,	O
the	O
granularity	O
of	O
prediction	O
types	O
is	O
brought	O
down	O
to	O
the	O
"	O
slice	O
level.	O
"	O
</s>
<s>
I-slices	O
,	O
P-slices	O
,	O
and	O
B-slices	O
take	O
the	O
place	O
of	O
I	O
,	O
P	O
,	O
and	O
B	B-General_Concept
frames	I-General_Concept
.	O
</s>
<s>
Typically	O
,	O
pictures	O
(	O
frames	O
)	O
are	O
segmented	O
into	O
macroblocks	B-General_Concept
,	O
and	O
individual	O
prediction	O
types	O
can	O
be	O
selected	O
on	O
a	O
macroblock	B-General_Concept
basis	O
rather	O
than	O
being	O
the	O
same	O
for	O
the	O
entire	O
picture	O
,	O
as	O
follows	O
:	O
</s>
<s>
Furthermore	O
,	O
in	O
the	O
H.264	B-Application
video	O
coding	O
standard	O
,	O
the	O
frame	O
can	O
be	O
segmented	O
into	O
sequences	O
of	O
macroblocks	B-General_Concept
called	O
slices	O
,	O
and	O
instead	O
of	O
using	O
I	O
,	O
B	O
and	O
P-frame	O
type	O
selections	O
,	O
the	O
encoder	O
can	O
choose	O
the	O
prediction	O
style	O
distinctly	O
on	O
each	O
individual	O
slice	O
.	O
</s>
<s>
Also	O
in	O
H.264	B-Application
are	O
found	O
several	O
additional	O
types	O
of	O
frames/slices	O
:	O
</s>
<s>
SI‑	O
frames/slices	O
(	O
Switching	O
I	O
)	O
:	O
Facilitates	O
switching	O
between	O
coded	O
streams	O
;	O
contains	O
SI-macroblocks	O
(	O
a	O
special	O
type	O
of	O
intra	O
coded	O
macroblock	B-General_Concept
)	O
.	O
</s>
<s>
SI	O
and	O
SP	O
frames	O
(	O
defined	O
for	O
the	O
Extended	O
Profile	O
)	O
improve	O
error	B-Error_Name
correction	I-Error_Name
.	O
</s>
<s>
I-frames	B-General_Concept
contain	O
an	O
entire	O
image	O
.	O
</s>
<s>
May	O
also	O
be	O
generated	O
when	O
differentiating	O
image	O
details	O
prohibit	O
generation	O
of	O
effective	O
P	O
or	O
B-frames	B-General_Concept
.	O
</s>
<s>
Intra	O
refresh	O
periods	O
of	O
a	O
half-second	O
are	O
common	O
on	O
such	O
applications	O
as	O
digital	B-Application
television	I-Application
broadcast	O
and	O
DVD	O
storage	O
.	O
</s>
<s>
For	O
example	O
,	O
in	O
videoconferencing	B-Application
systems	O
it	O
is	O
common	O
to	O
send	O
I-frames	B-General_Concept
very	O
infrequently	O
.	O
</s>
<s>
Older	O
standard	O
designs	O
(	O
such	O
as	O
MPEG-2	B-Algorithm
)	O
use	O
only	O
one	O
previously	O
decoded	O
picture	O
as	O
a	O
reference	O
during	O
decoding	O
,	O
and	O
require	O
that	O
picture	O
to	O
also	O
precede	O
the	O
P	O
picture	O
in	O
display	O
order	O
.	O
</s>
<s>
In	O
H.264	B-Application
,	O
can	O
use	O
multiple	O
previously	O
decoded	O
pictures	O
as	O
references	O
during	O
decoding	O
,	O
and	O
can	O
have	O
any	O
arbitrary	O
display-order	O
relationship	O
relative	O
to	O
the	O
picture(s )	O
used	O
for	O
its	O
prediction	O
.	O
</s>
<s>
Older	O
standards	O
allow	O
only	O
a	O
single	O
global	B-Algorithm
motion	I-Algorithm
compensation	I-Algorithm
vector	O
for	O
the	O
entire	O
frame	O
or	O
a	O
single	O
motion	O
compensation	O
vector	O
per	O
macroblock	B-General_Concept
.	O
</s>
<s>
Include	O
some	O
prediction	O
modes	O
that	O
form	O
a	O
prediction	O
of	O
a	O
motion	O
region	O
(	O
e.g.	O
,	O
a	O
macroblock	B-General_Concept
or	O
a	O
smaller	O
area	O
)	O
by	O
averaging	O
the	O
predictions	O
obtained	O
using	O
two	O
different	O
previously	O
decoded	O
reference	O
regions	O
.	O
</s>
<s>
Some	O
standards	O
allow	O
two	O
motion	O
compensation	O
vectors	O
per	O
macroblock	B-General_Concept
(	O
biprediction	O
)	O
.	O
</s>
<s>
In	O
older	O
standards	O
(	O
such	O
as	O
MPEG-2	B-Algorithm
)	O
,	O
B-frames	B-General_Concept
are	O
never	O
used	O
as	O
references	O
for	O
the	O
prediction	O
of	O
other	O
pictures	O
.	O
</s>
<s>
As	O
a	O
result	O
,	O
a	O
lower	O
quality	O
encoding	O
(	O
requiring	O
less	O
space	O
)	O
can	O
be	O
used	O
for	O
such	O
B-frames	B-General_Concept
because	O
the	O
loss	O
of	O
detail	O
will	O
not	O
harm	O
the	O
prediction	O
quality	O
for	O
subsequent	O
pictures	O
.	O
</s>
<s>
H.264	B-Application
relaxes	O
this	O
restriction	O
,	O
and	O
allows	O
B-frames	B-General_Concept
to	O
be	O
used	O
as	O
references	O
for	O
the	O
decoding	O
of	O
other	O
frames	O
at	O
the	O
encoder	O
's	O
discretion	O
.	O
</s>
<s>
Older	O
standards	O
(	O
such	O
as	O
MPEG-2	B-Algorithm
)	O
,	O
use	O
exactly	O
two	O
previously	O
decoded	O
pictures	O
as	O
references	O
during	O
decoding	O
,	O
and	O
require	O
one	O
of	O
those	O
pictures	O
to	O
precede	O
the	O
B-frame	O
in	O
display	O
order	O
and	O
the	O
other	O
one	O
to	O
follow	O
it	O
.	O
</s>
<s>
H.264	B-Application
allows	O
for	O
one	O
,	O
two	O
,	O
or	O
more	O
than	O
two	O
previously	O
decoded	O
pictures	O
as	O
references	O
during	O
decoding	O
,	O
and	O
can	O
have	O
any	O
arbitrary	O
display-order	O
relationship	O
relative	O
to	O
the	O
picture(s )	O
used	O
for	O
its	O
prediction	O
.	O
</s>
<s>
The	O
heightened	O
flexibility	O
of	O
information	O
retrieval	O
means	O
that	O
B-frames	B-General_Concept
typically	O
require	O
fewer	O
bits	O
for	O
encoding	O
than	O
either	O
I	O
or	O
P-frames	O
.	O
</s>
