<s>
2D	O
to	O
3D	B-Algorithm
video	I-Algorithm
conversion	O
(	O
also	O
called	O
2D	O
to	O
stereo	B-Algorithm
3D	I-Algorithm
conversion	O
and	O
stereo	B-Algorithm
conversion	I-Algorithm
)	O
is	O
the	O
process	O
of	O
transforming	O
2D	O
(	O
"	O
flat	O
"	O
)	O
film	O
to	O
3D	B-Algorithm
form	O
,	O
which	O
in	O
almost	O
all	O
cases	O
is	O
stereo	B-Algorithm
,	O
so	O
it	O
is	O
the	O
process	O
of	O
creating	O
imagery	O
for	O
each	O
eye	O
from	O
one	O
2D	O
image	O
.	O
</s>
<s>
2D-to-3D	O
conversion	O
adds	O
the	O
binocular	O
disparity	O
depth	O
cue	O
to	O
digital	O
images	O
perceived	O
by	O
the	O
brain	O
,	O
thus	O
,	O
if	O
done	O
properly	O
,	O
greatly	O
improving	O
the	O
immersive	O
effect	O
while	O
viewing	O
stereo	B-Algorithm
video	O
in	O
comparison	O
to	O
2D	O
video	O
.	O
</s>
<s>
If	O
done	O
properly	O
and	O
thoroughly	O
,	O
the	O
conversion	O
produces	O
stereo	B-Algorithm
video	O
of	O
similar	O
quality	O
to	O
"	O
native	O
"	O
stereo	B-Algorithm
video	O
which	O
is	O
shot	O
in	O
stereo	B-Algorithm
and	O
accurately	O
adjusted	O
and	O
aligned	O
in	O
post-production	O
.	O
</s>
<s>
Two	O
approaches	O
to	O
stereo	B-Algorithm
conversion	I-Algorithm
can	O
be	O
loosely	O
defined	O
:	O
quality	O
semiautomatic	O
conversion	O
for	O
cinema	O
and	O
high	O
quality	O
3DTV	B-Algorithm
,	O
and	O
low-quality	O
automatic	O
conversion	O
for	O
cheap	O
3DTV	B-Algorithm
,	O
VOD	B-Application
and	O
similar	O
applications	O
.	O
</s>
<s>
Computer	O
animated	O
2D	O
films	O
made	O
with	O
3D	B-Algorithm
models	O
can	O
be	O
re-rendered	O
in	O
stereoscopic	B-Algorithm
3D	I-Algorithm
by	O
adding	O
a	O
second	O
virtual	O
camera	O
if	O
the	O
original	O
data	O
is	O
still	O
available	O
.	O
</s>
<s>
This	O
is	O
technically	O
not	O
a	O
conversion	O
;	O
therefore	O
,	O
such	O
re-rendered	O
films	O
have	O
the	O
same	O
quality	O
as	O
films	O
originally	O
produced	O
in	O
stereoscopic	B-Algorithm
3D	I-Algorithm
.	O
</s>
<s>
Examples	O
of	O
this	O
technique	O
include	O
the	O
re-release	O
of	O
Toy	B-Application
Story	I-Application
and	O
Toy	B-Application
Story	I-Application
2	I-Application
.	O
</s>
<s>
Revisiting	O
the	O
original	O
computer	O
data	O
for	O
the	O
two	O
films	O
took	O
four	O
months	O
,	O
as	O
well	O
as	O
an	O
additional	O
six	O
months	O
to	O
add	O
the	O
3D	B-Algorithm
.	O
</s>
<s>
However	O
,	O
not	O
all	O
CGI	O
films	O
are	O
re-rendered	O
for	O
the	O
3D	B-Algorithm
re-release	O
because	O
of	O
the	O
costs	O
,	O
time	O
required	O
,	O
lack	O
of	O
skilled	O
resources	O
or	O
missing	O
computer	O
data	O
.	O
</s>
<s>
With	O
the	O
increase	O
of	O
films	O
released	O
in	O
3D	B-Algorithm
,	O
2D	B-Algorithm
to	I-Algorithm
3D	I-Algorithm
conversion	I-Algorithm
has	O
become	O
more	O
common	O
.	O
</s>
<s>
The	O
majority	O
of	O
non-CGI	O
stereo	B-Algorithm
3D	I-Algorithm
blockbusters	O
are	O
converted	O
fully	O
or	O
at	O
least	O
partially	O
from	O
2D	O
footage	O
.	O
</s>
<s>
Even	O
Avatar	O
contains	O
several	O
scenes	O
shot	O
in	O
2D	O
and	O
converted	O
to	O
stereo	B-Algorithm
in	O
post-production	O
.	O
</s>
<s>
The	O
reasons	O
for	O
shooting	O
in	O
2D	O
instead	O
of	O
stereo	B-Algorithm
are	O
financial	O
,	O
technical	O
and	O
sometimes	O
artistic	O
:	O
</s>
<s>
Stereo	B-Algorithm
post-production	O
workflow	O
is	O
much	O
more	O
complex	O
and	O
not	O
as	O
well-established	O
as	O
2D	O
workflow	O
,	O
requiring	O
more	O
work	O
and	O
rendering	O
.	O
</s>
<s>
Professional	O
stereoscopic	B-Algorithm
rigs	O
are	O
much	O
more	O
expensive	O
and	O
bulky	O
than	O
customary	O
monocular	O
cameras	O
.	O
</s>
<s>
Stereo	B-Algorithm
cameras	O
can	O
introduce	O
various	O
mismatches	O
in	O
stereo	B-Algorithm
image	O
(	O
such	O
as	O
vertical	O
parallax	B-Device
,	O
tilt	O
,	O
color	O
shift	O
,	O
reflections	O
and	O
glares	O
in	O
different	O
positions	O
)	O
that	O
should	O
be	O
fixed	O
in	O
post-production	O
anyway	O
because	O
they	O
ruin	O
the	O
3D	B-Algorithm
effect	O
.	O
</s>
<s>
This	O
correction	O
sometimes	O
may	O
have	O
complexity	O
comparable	O
to	O
stereo	B-Algorithm
conversion	I-Algorithm
.	O
</s>
<s>
Stereo	B-Algorithm
cameras	O
can	O
betray	O
practical	O
effects	O
used	O
during	O
filming	O
.	O
</s>
<s>
The	O
same	O
scene	O
filmed	O
in	O
stereo	B-Algorithm
would	O
reveal	O
that	O
the	O
actors	O
were	O
not	O
the	O
same	O
distance	O
from	O
the	O
camera	O
.	O
</s>
<s>
By	O
their	O
very	O
nature	O
,	O
stereo	B-Algorithm
cameras	O
have	O
restrictions	O
on	O
how	O
far	O
the	O
camera	O
can	O
be	O
from	O
the	O
filmed	O
subject	O
and	O
still	O
provide	O
acceptable	O
stereo	B-Algorithm
separation	O
.	O
</s>
<s>
However	O
,	O
while	O
the	O
zoom	O
lens	O
would	O
provide	O
acceptable	O
image	O
quality	O
,	O
the	O
stereo	B-Algorithm
separation	O
would	O
be	O
virtually	O
nil	O
over	O
such	O
a	O
distance	O
.	O
</s>
<s>
Even	O
in	O
the	O
case	O
of	O
stereo	B-Algorithm
shooting	O
,	O
conversion	O
can	O
frequently	O
be	O
necessary	O
.	O
</s>
<s>
Besides	O
the	O
mentioned	O
hard-to-shoot	O
scenes	O
,	O
there	O
are	O
situations	O
when	O
mismatches	O
in	O
stereo	B-Algorithm
views	O
are	O
too	O
big	O
to	O
adjust	O
,	O
and	O
it	O
is	O
simpler	O
to	O
perform	O
2D	O
to	O
stereo	B-Algorithm
conversion	I-Algorithm
,	O
treating	O
one	O
of	O
the	O
views	O
as	O
the	O
original	O
2D	O
source	O
.	O
</s>
<s>
If	O
an	O
object	O
in	O
stereo	B-Algorithm
pair	O
is	O
in	O
exactly	O
the	O
same	O
spot	O
for	O
both	O
eyes	O
,	O
then	O
it	O
will	O
appear	O
on	O
the	O
screen	O
surface	O
and	O
it	O
will	O
be	O
in	O
zero	O
parallax	B-Device
.	O
</s>
<s>
Objects	O
in	O
front	O
of	O
the	O
screen	O
are	O
said	O
to	O
be	O
in	O
negative	O
parallax	B-Device
,	O
and	O
background	O
imagery	O
behind	O
the	O
screen	O
is	O
in	O
positive	O
parallax	B-Device
.	O
</s>
<s>
Filling	O
of	O
uncovered	O
areas	O
–	O
left	O
or	O
right	O
view	O
images	O
show	O
a	O
scene	O
from	O
a	O
different	O
angle	O
,	O
and	O
parts	O
of	O
objects	O
or	O
entire	O
objects	O
covered	O
by	O
the	O
foreground	O
in	O
the	O
original	O
2D	O
image	O
should	O
become	O
visible	O
in	O
a	O
stereo	B-Algorithm
pair	O
.	O
</s>
<s>
Otherwise	O
the	O
unknown	O
areas	O
must	O
be	O
filled	O
in	O
by	O
an	O
artist	O
or	O
inpainted	B-Algorithm
,	O
since	O
the	O
exact	O
reconstruction	O
is	O
not	O
possible	O
.	O
</s>
<s>
Most	O
semiautomatic	O
methods	O
of	O
stereo	B-Algorithm
conversion	I-Algorithm
use	O
depth	O
maps	O
and	O
depth-image-based	O
rendering	O
.	O
</s>
<s>
While	O
depth	O
mapping	O
can	O
produce	O
a	O
fairly	O
potent	O
illusion	O
of	O
3D	B-Algorithm
objects	O
in	O
the	O
video	O
,	O
it	O
inherently	O
does	O
not	O
support	O
semi-transparent	O
objects	O
or	O
areas	O
,	O
nor	O
does	O
it	O
represent	O
occluded	O
surfaces	O
;	O
to	O
emphasize	O
this	O
limitation	O
,	O
depth-based	O
3D	B-Algorithm
representations	O
are	O
often	O
explicitly	O
referred	O
to	O
as	O
2.5D	O
.	O
</s>
<s>
Image	B-Algorithm
segmentation	I-Algorithm
,	O
creation	O
of	O
mattes	O
or	O
masks	O
,	O
usually	O
by	O
rotoscoping	O
.	O
</s>
<s>
This	O
is	O
an	O
iterative	O
process	O
requiring	O
adjustment	O
of	O
objects	O
,	O
shapes	O
,	O
depth	O
,	O
and	O
visualization	O
of	O
intermediate	O
results	O
in	O
stereo	B-Algorithm
.	O
</s>
<s>
Depth	O
micro-relief	O
,	O
3D	B-Algorithm
shape	O
is	O
added	O
to	O
most	O
important	O
surfaces	O
to	O
prevent	O
the	O
"	O
cardboard	O
"	O
effect	O
when	O
stereo	B-Algorithm
imagery	O
looks	O
like	O
a	O
combination	O
of	O
flat	O
images	O
just	O
set	O
at	O
different	O
depths	O
.	O
</s>
<s>
Stereo	B-Algorithm
generation	O
based	O
on	O
2D+Depth	B-Algorithm
with	O
any	O
supplemental	O
information	O
like	O
clean	O
plates	O
,	O
restored	O
background	O
,	O
transparency	O
maps	O
,	O
etc	O
.	O
</s>
<s>
Usually	O
the	O
original	O
2D	O
image	O
is	O
treated	O
as	O
the	O
center	O
image	O
,	O
so	O
that	O
two	O
stereo	B-Algorithm
views	O
are	O
generated	O
.	O
</s>
<s>
During	O
stereo	B-Algorithm
generation	O
,	O
pixels	O
of	O
the	O
original	O
image	O
are	O
shifted	O
to	O
the	O
left	O
or	O
to	O
the	O
right	O
depending	O
on	O
depth	O
map	O
,	O
maximum	O
selected	O
parallax	B-Device
,	O
and	O
screen	O
surface	O
position	O
.	O
</s>
<s>
Reconstruction	O
and	O
painting	O
of	O
any	O
uncovered	O
areas	O
not	O
filled	O
by	O
the	O
stereo	B-Algorithm
generator	O
.	O
</s>
<s>
Stereo	B-Algorithm
can	O
be	O
presented	O
in	O
any	O
format	O
for	O
preview	O
purposes	O
,	O
including	O
anaglyph	B-Algorithm
.	O
</s>
<s>
The	O
more	O
layers	O
are	O
processed	O
separately	O
per	O
frame	O
,	O
the	O
higher	O
the	O
quality	O
of	O
3D	B-Algorithm
illusion	O
tends	O
to	O
be	O
.	O
</s>
<s>
3D	B-Algorithm
reconstruction	I-Algorithm
and	O
re-projection	O
may	O
be	O
used	O
for	O
stereo	B-Algorithm
conversion	I-Algorithm
.	O
</s>
<s>
It	O
involves	O
scene	O
3D	B-Algorithm
model	O
creation	O
,	O
extraction	O
of	O
original	O
image	O
surfaces	O
as	O
textures	O
for	O
3D	B-Algorithm
objects	O
and	O
,	O
finally	O
,	O
rendering	O
the	O
3D	B-Algorithm
scene	O
from	O
two	O
virtual	O
cameras	O
to	O
acquire	O
stereo	B-Algorithm
video	O
.	O
</s>
<s>
On	O
"	O
depth	O
from	O
defocus	O
"	O
(	O
DFD	O
)	O
approaches	O
,	O
the	O
depth	O
information	O
is	O
estimated	O
based	O
on	O
the	O
amount	O
of	O
blur	O
of	O
the	O
considered	O
object	O
,	O
whereas	O
"	O
depth	O
from	O
focus	O
"	O
(	O
DFF	O
)	O
approaches	O
tend	O
to	O
compare	O
the	O
sharpness	O
of	O
an	O
object	O
over	O
a	O
range	O
of	O
images	O
taken	O
with	O
different	O
focus	B-Algorithm
distances	I-Algorithm
in	O
order	O
to	O
find	O
out	O
its	O
distance	O
to	O
the	O
camera	O
.	O
</s>
<s>
HV3D	O
quality	O
metric	O
has	O
been	O
designed	O
having	O
the	O
human	O
visual	O
3D	B-Algorithm
perception	O
in	O
mind	O
.	O
</s>
