<s>
A	O
Guided	B-Algorithm
filter	I-Algorithm
is	O
an	O
edge-preserving	B-Algorithm
smoothing	I-Algorithm
light	O
filter	O
.	O
</s>
<s>
As	O
with	O
a	O
bilateral	B-Algorithm
filter	I-Algorithm
,	O
it	O
can	O
filter	O
out	O
noise	O
or	O
texture	O
while	O
retaining	O
sharp	O
edges	O
.	O
</s>
<s>
Compared	O
to	O
the	O
bilateral	B-Algorithm
filter	I-Algorithm
,	O
the	O
guided	O
image	O
filter	O
has	O
two	O
advantages	O
:	O
bilateral	B-Algorithm
filters	I-Algorithm
have	O
high	O
computational	O
complexity	O
,	O
while	O
the	O
guided	O
image	O
filter	O
uses	O
simpler	O
calculations	O
with	O
linear	O
computational	O
complexity	O
.	O
</s>
<s>
Bilateral	B-Algorithm
filters	I-Algorithm
sometimes	O
include	O
unwanted	O
gradient	O
reversal	O
artifacts	O
and	O
cause	O
image	O
distortion	O
.	O
</s>
<s>
One	O
key	O
assumption	O
of	O
the	O
guided	B-Algorithm
filter	I-Algorithm
is	O
that	O
the	O
relation	O
between	O
guidance	O
and	O
the	O
filtering	O
output	O
is	O
linear	O
.	O
</s>
<s>
The	O
guided	B-Algorithm
filter	I-Algorithm
removes	O
noise	O
in	O
the	O
input	O
image	O
while	O
preserving	O
clear	O
edges	O
.	O
</s>
<s>
Specifically	O
,	O
a	O
“	O
flat	O
patch	O
”	O
or	O
a	O
“	O
high	O
variance	O
patch	O
”	O
can	O
be	O
specified	O
by	O
the	O
parameter	O
of	O
the	O
guided	B-Algorithm
filter	I-Algorithm
.	O
</s>
<s>
The	O
role	O
of	O
the	O
range	O
variance	O
in	O
the	O
bilateral	B-Algorithm
filter	I-Algorithm
is	O
similar	O
to	O
in	O
the	O
guided	B-Algorithm
filter	I-Algorithm
.	O
</s>
<s>
When	O
using	O
the	O
bilateral	B-Algorithm
filter	I-Algorithm
to	O
filter	O
an	O
image	O
,	O
artifacts	O
may	O
appear	O
on	O
the	O
edges	O
.	O
</s>
<s>
The	O
guided	B-Algorithm
filter	I-Algorithm
performs	O
better	O
in	O
avoiding	O
gradient	O
reversal	O
.	O
</s>
