<s>
An	O
inter	B-Algorithm
frame	I-Algorithm
is	O
a	O
frame	O
in	O
a	O
video	O
compression	O
stream	O
which	O
is	O
expressed	O
in	O
terms	O
of	O
one	O
or	O
more	O
neighboring	O
frames	O
.	O
</s>
<s>
The	O
"	O
inter	O
"	O
part	O
of	O
the	O
term	O
refers	O
to	O
the	O
use	O
of	O
Inter	B-Algorithm
frame	I-Algorithm
prediction	I-Algorithm
.	O
</s>
<s>
An	O
inter	O
coded	O
frame	O
is	O
divided	O
into	O
blocks	O
known	O
as	O
macroblocks	B-General_Concept
.	O
</s>
<s>
After	O
that	O
,	O
instead	O
of	O
directly	O
encoding	O
the	O
raw	O
pixel	O
values	O
for	O
each	O
block	O
,	O
the	O
encoder	O
will	O
try	O
to	O
find	O
a	O
block	O
similar	O
to	O
the	O
one	O
it	O
is	O
encoding	O
on	O
a	O
previously	O
encoded	O
frame	O
,	O
referred	O
to	O
as	O
a	O
reference	B-General_Concept
frame	I-General_Concept
.	O
</s>
<s>
This	O
process	O
is	O
done	O
by	O
a	O
block	B-General_Concept
matching	I-General_Concept
algorithm	I-General_Concept
.	O
</s>
<s>
If	O
the	O
encoder	O
succeeds	O
on	O
its	O
search	O
,	O
the	O
block	O
could	O
be	O
encoded	O
by	O
a	O
vector	O
,	O
known	O
as	O
motion	O
vector	O
,	O
which	O
points	O
to	O
the	O
position	O
of	O
the	O
matching	O
block	O
at	O
the	O
reference	B-General_Concept
frame	I-General_Concept
.	O
</s>
<s>
To	O
sum	O
up	O
,	O
if	O
the	O
encoder	O
succeeds	O
in	O
finding	O
a	O
matching	O
block	O
on	O
a	O
reference	B-General_Concept
frame	I-General_Concept
,	O
it	O
will	O
obtain	O
a	O
motion	O
vector	O
pointing	O
to	O
the	O
matched	O
block	O
and	O
a	O
prediction	O
error	O
.	O
</s>
<s>
If	O
the	O
block	B-General_Concept
matching	I-General_Concept
algorithm	I-General_Concept
fails	O
to	O
find	O
a	O
suitable	O
match	O
the	O
prediction	O
error	O
will	O
be	O
considerable	O
.	O
</s>
<s>
If	O
the	O
matched	O
block	O
at	O
the	O
reference	B-General_Concept
frame	I-General_Concept
has	O
also	O
been	O
encoded	O
using	O
Inter	B-Algorithm
frame	I-Algorithm
prediction	I-Algorithm
,	O
the	O
errors	O
made	O
for	O
its	O
encoding	O
will	O
be	O
propagated	O
to	O
the	O
next	O
block	O
.	O
</s>
<s>
Because	O
of	O
these	O
drawbacks	O
,	O
a	O
reliable	O
and	O
time	O
periodic	O
reference	B-General_Concept
frame	I-General_Concept
must	O
be	O
used	O
for	O
this	O
technique	O
to	O
be	O
efficient	O
and	O
useful	O
.	O
</s>
<s>
That	O
reference	B-General_Concept
frame	I-General_Concept
is	O
known	O
as	O
Intra-frame	B-General_Concept
,	O
which	O
is	O
strictly	O
intra	O
coded	O
,	O
so	O
it	O
can	O
always	O
be	O
decoded	O
without	O
additional	O
information	O
.	O
</s>
<s>
In	O
most	O
designs	O
,	O
there	O
are	O
two	O
types	O
of	O
inter	B-Algorithm
frames	I-Algorithm
:	O
P-frames	O
and	O
B-frames	B-General_Concept
.	O
</s>
<s>
These	O
two	O
kinds	O
of	O
frames	O
and	O
the	O
I-frames	B-General_Concept
(	O
Intra-coded	O
pictures	O
)	O
usually	O
join	O
in	O
a	O
GOP	B-General_Concept
(	O
Group	B-General_Concept
Of	I-General_Concept
Pictures	I-General_Concept
)	O
.	O
</s>
<s>
The	O
I-frame	B-General_Concept
does	O
n't	O
need	O
additional	O
information	O
to	O
be	O
decoded	O
and	O
it	O
can	O
be	O
used	O
as	O
a	O
reliable	O
reference	O
.	O
</s>
<s>
This	O
structure	O
also	O
allows	O
to	O
achieve	O
an	O
I-frame	B-General_Concept
periodicity	O
,	O
which	O
is	O
needed	O
for	O
decoder	O
synchronization	O
.	O
</s>
<s>
The	O
difference	O
between	O
P-frames	O
and	O
B-frames	B-General_Concept
is	O
the	O
reference	B-General_Concept
frame	I-General_Concept
they	O
are	O
allowed	O
to	O
use	O
.	O
</s>
<s>
The	O
prediction	O
is	O
made	O
from	O
an	O
earlier	O
picture	O
,	O
mainly	O
an	O
I-frame	B-General_Concept
or	O
P-frame	O
,	O
so	O
that	O
require	O
less	O
coding	O
data	O
( ≈50%	O
when	O
compared	O
to	O
I-frame	B-General_Concept
size	O
)	O
.	O
</s>
<s>
It	O
involves	O
the	O
use	O
of	O
motion	B-Algorithm
compensation	I-Algorithm
.	O
</s>
<s>
This	O
kind	O
of	O
prediction	O
method	O
occupies	O
less	O
coding	O
data	O
than	O
P-frames	O
generally	O
( ≈25%	O
when	O
compared	O
to	O
I-frame	B-General_Concept
size	O
)	O
because	O
the	O
prediction	O
is	O
made	O
from	O
either	O
an	O
earlier	O
frame	O
or	O
a	O
later	O
frame	O
or	O
both	O
them	O
.	O
</s>
<s>
(	O
B-frames	B-General_Concept
can	O
also	O
be	O
less	O
efficient	O
than	O
P-frames	O
under	O
certain	O
cases	O
,	O
e.g.	O
</s>
<s>
Similar	O
to	O
P-frames	O
,	O
B-frames	B-General_Concept
are	O
expressed	O
as	O
motion	O
vectors	O
and	O
transform	O
coefficients	O
.	O
</s>
<s>
In	O
order	O
to	O
avoid	O
a	O
growing	O
propagation	O
error	O
,	O
B-frames	B-General_Concept
are	O
not	O
used	O
as	O
a	O
reference	O
to	O
make	O
further	O
predictions	O
in	O
most	O
encoding	O
standards	O
.	O
</s>
<s>
However	O
,	O
in	O
newer	O
encoding	O
methods	O
(	O
such	O
as	O
H.264/MPEG	B-Application
-4	I-Application
AVC	I-Application
and	O
HEVC	B-Algorithm
)	O
,	O
B-frames	B-General_Concept
may	O
be	O
used	O
as	O
reference	O
for	O
better	O
exploitation	O
of	O
temporal	O
redundancy	O
.	O
</s>
<s>
The	O
typical	O
Group	B-General_Concept
of	I-General_Concept
pictures	I-General_Concept
(	O
GOP	B-General_Concept
)	O
structure	O
is	O
IBBPBBP	O
...	O
</s>
<s>
The	O
I-frame	B-General_Concept
is	O
used	O
to	O
predict	O
the	O
first	O
P-frame	O
and	O
these	O
two	O
frames	O
are	O
also	O
used	O
to	O
predict	O
the	O
first	O
and	O
the	O
second	O
B-frames	B-General_Concept
.	O
</s>
<s>
The	O
second	O
P-frame	O
is	O
predicted	O
also	O
using	O
the	O
first	O
I-frame	B-General_Concept
.	O
</s>
<s>
Both	O
P-frames	O
join	O
to	O
predict	O
the	O
third	O
and	O
fourth	O
B-frames	B-General_Concept
.	O
</s>
<s>
This	O
structure	O
suggests	O
a	O
problem	O
because	O
the	O
fourth	O
frame	O
(	O
a	O
P-frame	O
)	O
is	O
needed	O
in	O
order	O
to	O
predict	O
the	O
second	O
and	O
the	O
third	O
(	O
B-frames	B-General_Concept
)	O
.	O
</s>
<s>
So	O
we	O
need	O
to	O
transmit	O
the	O
P-frame	O
before	O
the	O
B-frames	B-General_Concept
and	O
it	O
will	O
delay	O
the	O
transmission	O
(	O
it	O
will	O
be	O
necessary	O
to	O
keep	O
the	O
P-frame	O
)	O
.	O
</s>
<s>
P-frames	O
and	O
B-frames	B-General_Concept
need	O
less	O
data	O
than	O
I-frames	B-General_Concept
,	O
so	O
less	O
data	O
is	O
transmitted	O
.	O
</s>
<s>
B-frames	B-General_Concept
may	O
introduce	O
decoding	O
dependency	O
which	O
inevitably	O
increases	O
decoding	O
latency	O
.	O
</s>
<s>
The	O
most	O
important	O
improvements	O
of	O
this	O
technique	O
in	O
regard	O
to	O
previous	O
H.264	B-Application
standard	O
are	O
:	O
</s>
<s>
Luminance	O
block	O
partition	O
of	O
16×16	O
(	O
MPEG-2	B-Algorithm
)	O
,	O
16×8	O
,	O
8×16	O
,	O
and	O
8×8	O
.	O
</s>
<s>
While	O
MPEG-2	B-Algorithm
allowed	O
a	O
½	O
pixel	O
resolution	O
,	O
Inter	B-Algorithm
frame	I-Algorithm
allows	O
up	O
to	O
¼	O
pixel	O
resolution	O
.	O
</s>
<s>
That	O
means	O
that	O
it	O
is	O
possible	O
to	O
search	O
a	O
block	O
in	O
the	O
frame	O
to	O
be	O
coded	O
in	O
other	O
reference	B-General_Concept
frames	I-General_Concept
,	O
or	O
we	O
can	O
interpolate	O
nonexistent	O
pixels	O
to	O
find	O
blocks	O
that	O
are	O
even	O
better	O
suited	O
to	O
the	O
current	O
block	O
.	O
</s>
<s>
Skip	O
and	O
Direct	O
Mode	O
are	O
very	O
frequently	O
used	O
,	O
especially	O
with	O
B-frames	B-General_Concept
.	O
</s>
<s>
The	O
encoder	O
will	O
only	O
record	O
that	O
it	O
is	O
a	O
Skip	O
Macroblock	B-General_Concept
.	O
</s>
<s>
Spatial	O
It	O
predicts	O
the	O
movement	O
from	O
neighbour	O
macroblocks	B-General_Concept
in	O
same	O
frame	O
.	O
</s>
<s>
As	O
we	O
can	O
see	O
,	O
they	O
are	O
used	O
very	O
frequently	O
,	O
mainly	O
in	O
B-frames	B-General_Concept
.	O
</s>
<s>
Although	O
the	O
use	O
of	O
the	O
term	O
"	O
frame	O
"	O
is	O
common	O
in	O
informal	O
usage	O
,	O
in	O
many	O
cases	O
(	O
such	O
as	O
in	O
international	O
standards	O
for	O
video	O
coding	O
by	O
MPEG	O
and	O
VCEG	B-Application
)	O
a	O
more	O
general	O
concept	O
is	O
applied	O
by	O
using	O
the	O
word	O
"	O
picture	O
"	O
rather	O
than	O
"	O
frame	O
"	O
,	O
where	O
a	O
picture	O
can	O
either	O
be	O
a	O
complete	O
frame	O
or	O
a	O
single	O
interlaced	O
field	O
.	O
</s>
<s>
Video	O
codecs	B-General_Concept
such	O
as	O
MPEG-2	B-Algorithm
,	O
H.264	B-Application
or	O
Ogg	B-Algorithm
Theora	I-Algorithm
reduce	O
the	O
amount	O
of	O
data	O
in	O
a	O
stream	O
by	O
following	O
key	O
frames	O
with	O
one	O
or	O
more	O
inter	B-Algorithm
frames	I-Algorithm
.	O
</s>
