<s>
Motion	B-Algorithm
compensation	I-Algorithm
in	O
computing	O
,	O
is	O
an	O
algorithmic	O
technique	O
used	O
to	O
predict	O
a	O
frame	O
in	O
a	O
video	O
,	O
given	O
the	O
previous	O
and/or	O
future	O
frames	O
by	O
accounting	O
for	O
motion	O
of	O
the	O
camera	O
and/or	O
objects	O
in	O
the	O
video	O
.	O
</s>
<s>
It	O
is	O
employed	O
in	O
the	O
encoding	O
of	O
video	O
data	O
for	O
video	O
compression	O
,	O
for	O
example	O
in	O
the	O
generation	O
of	O
MPEG-2	B-Algorithm
files	O
.	O
</s>
<s>
Motion	B-Algorithm
compensation	I-Algorithm
describes	O
a	O
picture	O
in	O
terms	O
of	O
the	O
transformation	O
of	O
a	O
reference	O
picture	O
to	O
the	O
current	O
picture	O
.	O
</s>
<s>
Motion	B-Algorithm
compensation	I-Algorithm
is	O
one	O
of	O
the	O
two	O
key	O
video	O
compression	B-General_Concept
techniques	I-General_Concept
used	O
in	O
video	B-Algorithm
coding	I-Algorithm
standards	I-Algorithm
,	O
along	O
with	O
the	O
discrete	B-General_Concept
cosine	I-General_Concept
transform	I-General_Concept
(	O
DCT	B-General_Concept
)	O
.	O
</s>
<s>
Most	O
video	B-Algorithm
coding	I-Algorithm
standards	I-Algorithm
,	O
such	O
as	O
the	O
H.26x	B-Application
and	O
MPEG	O
formats	O
,	O
typically	O
use	O
motion-compensated	O
DCT	B-General_Concept
hybrid	O
coding	O
,	O
known	O
as	O
block	O
motion	B-Algorithm
compensation	I-Algorithm
(	O
BMC	O
)	O
or	O
motion-compensated	O
DCT	B-General_Concept
(	O
MC	O
DCT	B-General_Concept
)	O
.	O
</s>
<s>
Motion	B-Algorithm
compensation	I-Algorithm
exploits	O
the	O
fact	O
that	O
,	O
often	O
,	O
for	O
many	O
frames	O
of	O
a	O
movie	O
,	O
the	O
only	O
difference	O
between	O
one	O
frame	O
and	O
another	O
is	O
the	O
result	O
of	O
either	O
the	O
camera	O
moving	O
or	O
an	O
object	O
in	O
the	O
frame	O
moving	O
.	O
</s>
<s>
Using	O
motion	B-Algorithm
compensation	I-Algorithm
,	O
a	O
video	O
stream	O
will	O
contain	O
some	O
full	O
(	O
reference	O
)	O
frames	O
;	O
then	O
the	O
only	O
information	O
stored	O
for	O
the	O
frames	O
in	O
between	O
would	O
be	O
the	O
information	O
needed	O
to	O
transform	O
the	O
previous	O
frame	O
into	O
the	O
next	O
frame	O
.	O
</s>
<s>
The	O
following	O
is	O
a	O
simplistic	O
illustrated	O
explanation	O
of	O
how	O
motion	B-Algorithm
compensation	I-Algorithm
works	O
.	O
</s>
<s>
This	O
also	O
means	O
that	O
it	O
is	O
also	O
possible	O
to	O
encode	O
the	O
information	O
using	O
difference	O
image	O
at	O
a	O
cost	O
of	O
less	O
compression	O
efficiency	O
but	O
by	O
saving	O
coding	O
complexity	O
without	O
motion	O
compensated	O
coding	O
;	O
as	O
a	O
matter	O
of	O
fact	O
that	O
motion	O
compensated	O
coding	O
(	O
together	O
with	O
motion	O
estimation	O
,	O
motion	B-Algorithm
compensation	I-Algorithm
)	O
occupies	O
more	O
than	O
90%	O
of	O
encoding	O
complexity	O
.	O
</s>
<s>
Motion	O
compensated	O
difference	O
160px	O
Differences	O
between	O
the	O
original	O
frame	O
and	O
the	O
next	O
frame	O
,	O
shifted	O
right	O
by	O
2	O
pixels	B-Algorithm
.	O
</s>
<s>
After	O
predicting	O
frames	O
using	O
motion	B-Algorithm
compensation	I-Algorithm
,	O
the	O
coder	O
finds	O
the	O
residual	O
,	O
which	O
is	O
then	O
compressed	O
and	O
transmitted	O
.	O
</s>
<s>
In	O
global	B-Algorithm
motion	I-Algorithm
compensation	I-Algorithm
,	O
the	O
motion	O
model	O
basically	O
reflects	O
camera	O
motions	O
such	O
as	O
:	O
</s>
<s>
There	O
are	O
several	O
advantages	O
of	O
global	B-Algorithm
motion	I-Algorithm
compensation	I-Algorithm
:	O
</s>
<s>
A	O
straight	O
line	O
(	O
in	O
the	O
time	O
direction	O
)	O
of	O
pixels	B-Algorithm
with	O
equal	O
spatial	O
positions	O
in	O
the	O
frame	O
corresponds	O
to	O
a	O
continuously	O
moving	O
point	O
in	O
the	O
real	O
scene	O
.	O
</s>
<s>
MPEG-4	B-Application
ASP	I-Application
supports	O
global	B-Algorithm
motion	I-Algorithm
compensation	I-Algorithm
with	O
three	O
reference	O
points	O
,	O
although	O
some	O
implementations	O
can	O
only	O
make	O
use	O
of	O
one	O
.	O
</s>
<s>
A	O
single	O
reference	O
point	O
only	O
allows	O
for	O
translational	O
motion	O
which	O
for	O
its	O
relatively	O
large	O
performance	O
cost	O
provides	O
little	O
advantage	O
over	O
block	O
based	O
motion	B-Algorithm
compensation	I-Algorithm
.	O
</s>
<s>
Moving	O
objects	O
within	O
a	O
frame	O
are	O
not	O
sufficiently	O
represented	O
by	O
global	B-Algorithm
motion	I-Algorithm
compensation	I-Algorithm
.	O
</s>
<s>
Block	O
motion	B-Algorithm
compensation	I-Algorithm
(	O
BMC	O
)	O
,	O
also	O
known	O
as	O
motion-compensated	O
discrete	B-General_Concept
cosine	I-General_Concept
transform	I-General_Concept
(	O
MC	O
DCT	B-General_Concept
)	O
,	O
is	O
the	O
most	O
widely	O
used	O
motion	B-Algorithm
compensation	I-Algorithm
technique	O
.	O
</s>
<s>
In	O
BMC	O
,	O
the	O
frames	O
are	O
partitioned	O
in	O
blocks	O
of	O
pixels	B-Algorithm
(	O
e.g.	O
</s>
<s>
macro-blocks	O
of	O
16×16	O
pixels	B-Algorithm
in	O
MPEG	O
)	O
.	O
</s>
<s>
The	O
result	O
of	O
this	O
differentiating	O
process	O
is	O
mathematically	O
equivalent	O
to	O
a	O
global	B-Algorithm
motion	I-Algorithm
compensation	I-Algorithm
capable	O
of	O
panning	O
.	O
</s>
<s>
Further	O
down	O
the	O
encoding	O
pipeline	O
,	O
an	O
entropy	B-Algorithm
coder	I-Algorithm
will	O
take	O
advantage	O
of	O
the	O
resulting	O
statistical	O
distribution	O
of	O
the	O
motion	O
vectors	O
around	O
the	O
zero	O
vector	O
to	O
reduce	O
the	O
output	O
size	O
.	O
</s>
<s>
It	O
is	O
possible	O
to	O
shift	O
a	O
block	O
by	O
a	O
non-integer	O
number	O
of	O
pixels	B-Algorithm
,	O
which	O
is	O
called	O
sub-pixel	O
precision	O
.	O
</s>
<s>
The	O
in-between	O
pixels	B-Algorithm
are	O
generated	O
by	O
interpolating	O
neighboring	O
pixels	B-Algorithm
.	O
</s>
<s>
Commonly	O
,	O
half-pixel	O
or	O
quarter	O
pixel	B-Algorithm
precision	O
(	O
Qpel	B-Algorithm
,	O
used	O
by	O
H.264	B-Application
and	O
MPEG-4/ASP	O
)	O
is	O
used	O
.	O
</s>
<s>
The	O
computational	O
expense	O
of	O
sub-pixel	O
precision	O
is	O
much	O
higher	O
due	O
to	O
the	O
extra	O
processing	O
required	O
for	O
interpolation	O
and	O
on	O
the	O
encoder	O
side	O
,	O
a	O
much	O
greater	O
number	O
of	O
potential	O
source	O
blocks	O
to	O
be	O
evaluated	O
.	O
</s>
<s>
The	O
main	O
disadvantage	O
of	O
block	O
motion	B-Algorithm
compensation	I-Algorithm
is	O
that	O
it	O
introduces	O
discontinuities	O
at	O
the	O
block	O
borders	O
(	O
blocking	O
artifacts	O
)	O
.	O
</s>
<s>
(	O
a	O
common	O
misconception	O
is	O
that	O
the	O
previous	O
frame	O
is	O
divided	O
up	O
into	O
non-overlapping	O
blocks	O
,	O
and	O
the	O
motion	B-Algorithm
compensation	I-Algorithm
vectors	O
tell	O
where	O
those	O
blocks	O
move	O
to	O
)	O
.	O
</s>
<s>
Some	O
video	O
compression	B-General_Concept
algorithms	I-General_Concept
assemble	O
the	O
current	O
frame	O
out	O
of	O
pieces	O
of	O
several	O
different	O
previously-transmitted	O
frames	O
.	O
</s>
<s>
These	O
bidirectionally	O
predicted	O
frames	O
are	O
called	O
B-frames	B-General_Concept
.	O
</s>
<s>
Further	O
,	O
the	O
use	O
of	O
triangular	O
tiles	O
has	O
also	O
been	O
proposed	O
for	O
motion	B-Algorithm
compensation	I-Algorithm
.	O
</s>
<s>
Under	O
this	O
scheme	O
,	O
the	O
frame	O
is	O
tiled	O
with	O
triangles	O
,	O
and	O
the	O
next	O
frame	O
is	O
generated	O
by	O
performing	O
an	O
affine	B-Algorithm
transformation	I-Algorithm
on	O
these	O
triangles	O
.	O
</s>
<s>
Only	O
the	O
affine	B-Algorithm
transformations	I-Algorithm
are	O
recorded/transmitted	O
.	O
</s>
<s>
Variable	O
block-size	O
motion	B-Algorithm
compensation	I-Algorithm
(	O
VBSMC	O
)	O
is	O
the	O
use	O
of	O
BMC	O
with	O
the	O
ability	O
for	O
the	O
encoder	O
to	O
dynamically	O
select	O
the	O
size	O
of	O
the	O
blocks	O
.	O
</s>
<s>
Other	O
areas	O
of	O
work	O
have	O
examined	O
the	O
use	O
of	O
variable-shape	O
feature	O
metrics	O
,	O
beyond	O
block	O
boundaries	O
,	O
from	O
which	O
interframe	B-Algorithm
vectors	O
can	O
be	O
calculated	O
.	O
</s>
<s>
Older	O
designs	O
such	O
as	O
H.261	B-Algorithm
and	O
MPEG-1	B-Algorithm
video	O
typically	O
use	O
a	O
fixed	O
block	O
size	O
,	O
while	O
newer	O
ones	O
such	O
as	O
H.263	B-Algorithm
,	O
MPEG-4	B-Application
Part	I-Application
2	I-Application
,	O
H.264/MPEG	B-Application
-4	I-Application
AVC	I-Application
,	O
and	O
VC-1	B-Operating_System
give	O
the	O
encoder	O
the	O
ability	O
to	O
dynamically	O
choose	O
what	O
block	O
size	O
will	O
be	O
used	O
to	O
represent	O
the	O
motion	O
.	O
</s>
<s>
Overlapped	O
block	O
motion	B-Algorithm
compensation	I-Algorithm
(	O
OBMC	O
)	O
is	O
a	O
good	O
solution	O
to	O
these	O
problems	O
because	O
it	O
not	O
only	O
increases	O
prediction	O
accuracy	O
but	O
also	O
avoids	O
blocking	O
artifacts	O
.	O
</s>
<s>
Thus	O
,	O
each	O
pixel	B-Algorithm
belongs	O
to	O
4	O
blocks	O
.	O
</s>
<s>
In	O
such	O
a	O
scheme	O
,	O
there	O
are	O
4	O
predictions	O
for	O
each	O
pixel	B-Algorithm
which	O
are	O
summed	O
up	O
to	O
a	O
weighted	O
mean	O
.	O
</s>
<s>
In	O
such	O
a	O
scheme	O
,	O
each	O
pixel	B-Algorithm
then	O
belongs	O
to	O
3	O
blocks	O
rather	O
than	O
4	O
,	O
and	O
rather	O
than	O
using	O
8	O
neighboring	O
blocks	O
,	O
only	O
4	O
are	O
used	O
for	O
each	O
block	O
to	O
be	O
compensated	O
.	O
</s>
<s>
In	O
motion	B-Algorithm
compensation	I-Algorithm
,	O
quarter	O
or	O
half	O
samples	O
are	O
actually	O
interpolated	O
sub-samples	O
caused	O
by	O
fractional	O
motion	O
vectors	O
.	O
</s>
<s>
See	O
subclause	O
8.4.2.2	O
"	O
Fractional	O
sample	O
interpolation	O
process	O
"	O
of	O
the	O
H.264	B-Application
standard	O
.	O
</s>
<s>
Still	O
image	O
coding	B-General_Concept
techniques	I-General_Concept
can	O
be	O
expanded	O
to	O
an	O
extra	O
dimension	O
.	O
</s>
<s>
Fractional	O
pixel	B-Algorithm
affine	B-Algorithm
transformations	I-Algorithm
lead	O
to	O
bleeding	O
between	O
adjacent	O
pixels	B-Algorithm
.	O
</s>
<s>
2D+Delta	B-Application
Encoding	O
techniques	O
utilize	O
H.264	B-Application
and	O
MPEG-2	B-Algorithm
compatible	O
coding	O
and	O
can	O
use	O
motion	B-Algorithm
compensation	I-Algorithm
to	O
compress	O
between	O
stereoscopic	O
images	O
.	O
</s>
<s>
A	O
precursor	O
to	O
the	O
concept	O
of	O
motion	B-Algorithm
compensation	I-Algorithm
dates	O
back	O
to	O
1929	O
,	O
when	O
R.D.	O
</s>
<s>
In	O
1959	O
,	O
the	O
concept	O
of	O
inter-frame	B-Algorithm
motion	B-Algorithm
compensation	I-Algorithm
was	O
proposed	O
by	O
NHK	O
researchers	O
Y	O
.	O
Taki	O
,	O
M	O
.	O
Hatori	O
and	O
S	O
.	O
Tanaka	O
,	O
who	O
proposed	O
predictive	O
inter-frame	B-Algorithm
video	B-Algorithm
coding	I-Algorithm
in	O
the	O
temporal	O
dimension	O
.	O
</s>
<s>
Practical	O
motion-compensated	O
video	O
compression	O
emerged	O
with	O
the	O
development	O
of	O
motion-compensated	O
DCT	B-General_Concept
(	O
MC	O
DCT	B-General_Concept
)	O
coding	O
,	O
also	O
called	O
block	O
motion	B-Algorithm
compensation	I-Algorithm
(	O
BMC	O
)	O
or	O
DCT	B-General_Concept
motion	B-Algorithm
compensation	I-Algorithm
.	O
</s>
<s>
This	O
is	O
a	O
hybrid	O
coding	O
algorithm	O
,	O
which	O
combines	O
two	O
key	O
data	B-General_Concept
compression	I-General_Concept
techniques	O
:	O
discrete	B-General_Concept
cosine	I-General_Concept
transform	I-General_Concept
(	O
DCT	B-General_Concept
)	O
coding	O
in	O
the	O
spatial	O
dimension	O
,	O
and	O
predictive	O
motion	B-Algorithm
compensation	I-Algorithm
in	O
the	O
temporal	O
dimension	O
.	O
</s>
<s>
DCT	B-General_Concept
coding	O
is	O
a	O
lossy	B-Algorithm
block	B-General_Concept
compression	I-General_Concept
transform	B-Algorithm
coding	I-Algorithm
technique	O
that	O
was	O
first	O
proposed	O
by	O
Nasir	O
Ahmed	O
,	O
who	O
initially	O
intended	O
it	O
for	O
image	B-General_Concept
compression	I-General_Concept
,	O
in	O
1972	O
.	O
</s>
<s>
In	O
1974	O
,	O
Ali	O
Habibi	O
at	O
the	O
University	O
of	O
Southern	O
California	O
introduced	O
hybrid	O
coding	O
,	O
which	O
combines	O
predictive	O
coding	O
with	O
transform	B-Algorithm
coding	I-Algorithm
.	O
</s>
<s>
However	O
,	O
his	O
algorithm	O
was	O
initially	O
limited	O
to	O
intra-frame	B-General_Concept
coding	I-General_Concept
in	O
the	O
spatial	O
dimension	O
.	O
</s>
<s>
In	O
1975	O
,	O
John	O
A	O
.	O
Roese	O
and	O
Guner	O
S	O
.	O
Robinson	O
extended	O
Habibi	O
's	O
hybrid	O
coding	O
algorithm	O
to	O
the	O
temporal	O
dimension	O
,	O
using	O
transform	B-Algorithm
coding	I-Algorithm
in	O
the	O
spatial	O
dimension	O
and	O
predictive	O
coding	O
in	O
the	O
temporal	O
dimension	O
,	O
developing	O
inter-frame	B-Algorithm
motion-compensated	O
hybrid	O
coding	O
.	O
</s>
<s>
For	O
the	O
spatial	O
transform	B-Algorithm
coding	I-Algorithm
,	O
they	O
experimented	O
with	O
the	O
DCT	B-General_Concept
and	O
the	O
fast	O
Fourier	O
transform	O
(	O
FFT	O
)	O
,	O
developing	O
inter-frame	B-Algorithm
hybrid	O
coders	O
for	O
both	O
,	O
and	O
found	O
that	O
the	O
DCT	B-General_Concept
is	O
the	O
most	O
efficient	O
due	O
to	O
its	O
reduced	O
complexity	O
,	O
capable	O
of	O
compressing	O
image	O
data	O
down	O
to	O
0.25-bit	O
per	O
pixel	B-Algorithm
for	O
a	O
videotelephone	B-Application
scene	O
with	O
image	O
quality	O
comparable	O
to	O
an	O
intra-frame	B-General_Concept
coder	O
requiring	O
2-bit	O
per	O
pixel	B-Algorithm
.	O
</s>
<s>
In	O
1977	O
,	O
Wen-Hsiung	O
Chen	O
developed	O
a	O
fast	O
DCT	B-General_Concept
algorithm	O
with	O
C.H.	O
</s>
<s>
In	O
1979	O
,	O
Anil	O
K	O
.	O
Jain	O
and	O
Jaswant	O
R	O
.	O
Jain	O
further	O
developed	O
motion-compensated	O
DCT	B-General_Concept
video	O
compression	O
,	O
also	O
called	O
block	O
motion	B-Algorithm
compensation	I-Algorithm
.	O
</s>
<s>
This	O
led	O
to	O
Chen	O
developing	O
a	O
practical	O
video	O
compression	B-General_Concept
algorithm	I-General_Concept
,	O
called	O
motion-compensated	O
DCT	B-General_Concept
or	O
adaptive	O
scene	O
coding	O
,	O
in	O
1981	O
.	O
</s>
<s>
Motion-compensated	O
DCT	B-General_Concept
later	O
became	O
the	O
standard	O
coding	O
technique	O
for	O
video	O
compression	O
from	O
the	O
late	O
1980s	O
onwards	O
.	O
</s>
<s>
The	O
first	O
digital	O
video	B-Algorithm
coding	I-Algorithm
standard	I-Algorithm
was	O
H.120	B-Algorithm
,	O
developed	O
by	O
the	O
CCITT	O
(	O
now	O
ITU-T	O
)	O
in	O
1984	O
.	O
</s>
<s>
H.120	B-Algorithm
used	O
motion-compensated	O
DPCM	O
coding	O
,	O
which	O
was	O
inefficient	O
for	O
video	B-Algorithm
coding	I-Algorithm
,	O
and	O
H.120	B-Algorithm
was	O
thus	O
impractical	O
due	O
to	O
low	O
performance	O
.	O
</s>
<s>
The	O
H.261	B-Algorithm
standard	I-Algorithm
was	O
developed	O
in	O
1988	O
based	O
on	O
motion-compensated	O
DCT	B-General_Concept
compression	O
,	O
and	O
it	O
was	O
the	O
first	O
practical	O
video	B-Algorithm
coding	I-Algorithm
standard	I-Algorithm
.	O
</s>
<s>
Since	O
then	O
,	O
motion-compensated	O
DCT	B-General_Concept
compression	O
has	O
been	O
adopted	O
by	O
all	O
the	O
major	O
video	B-Algorithm
coding	I-Algorithm
standards	I-Algorithm
(	O
including	O
the	O
H.26x	B-Application
and	O
MPEG	O
formats	O
)	O
that	O
followed	O
.	O
</s>
