<s>
Ordered	B-Algorithm
dithering	I-Algorithm
is	O
an	O
image	O
dithering	O
algorithm	O
.	O
</s>
<s>
For	O
example	O
,	O
Microsoft	B-Application
Windows	I-Application
uses	O
it	O
in	O
16-color	O
graphics	O
modes	O
.	O
</s>
<s>
This	O
threshold	O
map	O
(	O
for	O
sides	O
with	O
length	O
as	O
power	O
of	O
two	O
)	O
is	O
also	O
known	O
as	O
an	O
index	B-Algorithm
matrix	I-Algorithm
or	O
Bayer	O
matrix	O
.	O
</s>
<s>
The	O
ordered	B-Algorithm
dithering	I-Algorithm
algorithm	O
renders	O
the	O
image	O
normally	O
,	O
but	O
for	O
each	O
pixel	O
,	O
it	O
offsets	O
its	O
color	O
value	O
with	O
a	O
corresponding	O
value	O
from	O
the	O
threshold	O
map	O
according	O
to	O
its	O
location	O
,	O
causing	O
the	O
pixel	O
's	O
value	O
to	O
be	O
quantized	O
to	O
a	O
different	O
color	O
if	O
it	O
exceeds	O
the	O
threshold	O
.	O
</s>
<s>
Because	O
the	O
patterns	O
are	O
more	O
repetitive	O
than	O
error-diffusion	O
method	O
,	O
an	O
image	O
with	O
ordered	B-Algorithm
dithering	I-Algorithm
compresses	O
better	O
.	O
</s>
<s>
Ordered	B-Algorithm
dithering	I-Algorithm
is	O
more	O
suitable	O
for	O
line-art	O
graphics	O
as	O
it	O
will	O
result	O
in	O
straighter	O
lines	O
and	O
fewer	O
anomalies	O
.	O
</s>
<s>
For	O
example	O
,	O
when	O
quantizing	B-Algorithm
a	O
24	O
bpp	O
image	O
to	O
15	O
bpp	O
(	O
256	O
colors	O
per	O
channel	O
to	O
32	O
colors	O
per	O
channel	O
)	O
,	O
the	O
smallest	O
map	O
one	O
would	O
choose	O
would	O
be	O
4×2	O
,	O
for	O
the	O
ratio	O
of	O
8	O
(	O
256:32	O
)	O
.	O
</s>
<s>
The	O
above	O
thresholding	O
matrix	O
approach	O
describes	O
the	O
Bayer	O
family	O
of	O
ordered	B-Algorithm
dithering	I-Algorithm
algorithms	O
.	O
</s>
<s>
The	O
"	O
voids-and-cluster	O
"	O
method	O
gets	O
its	O
name	O
from	O
the	O
matrix	O
generation	O
procedure	O
,	O
where	O
a	O
black	O
image	O
with	O
randomly	O
initialized	O
white	O
pixels	O
is	O
gaussian-blurred	B-Error_Name
to	O
find	O
the	O
brightest	O
and	O
darkest	O
parts	O
,	O
corresponding	O
to	O
voids	O
and	O
clusters	O
.	O
</s>
