<s>
Color	B-Algorithm
normalization	I-Algorithm
is	O
a	O
topic	O
in	O
computer	B-Application
vision	I-Application
concerned	O
with	O
artificial	O
color	O
vision	O
and	O
object	O
recognition	O
.	O
</s>
<s>
Color	B-Algorithm
normalization	I-Algorithm
allows	O
for	O
object	O
recognition	O
techniques	O
based	O
on	O
color	O
to	O
compensate	O
for	O
these	O
variations	O
.	O
</s>
<s>
Color	B-Algorithm
normalization	I-Algorithm
has	O
been	O
used	O
for	O
object	O
recognition	O
on	O
color	O
images	O
in	O
the	O
field	O
of	O
robotics	O
,	O
bioinformatics	O
and	O
general	O
artificial	B-Application
intelligence	I-Application
,	O
when	O
it	O
is	O
important	O
to	O
remove	O
all	O
intensity	O
values	O
from	O
the	O
image	O
while	O
preserving	O
color	O
values	O
.	O
</s>
<s>
Another	O
area	O
where	O
color	B-Algorithm
normalization	I-Algorithm
is	O
commonly	O
used	O
is	O
in	O
pathology	O
,	O
and	O
more	O
specifically	O
,	O
digital	O
pathology	O
where	O
digitized	O
microscopic	O
images	O
of	O
histological	O
sections	O
can	O
be	O
stained	O
,	O
one	O
of	O
the	O
most	O
common	O
procedures	O
is	O
H&E	O
stain	O
.	O
</s>
<s>
The	O
main	O
issue	O
about	O
certain	O
applications	O
of	O
color	B-Algorithm
normalization	I-Algorithm
is	O
that	O
the	O
end	O
result	O
looks	O
unnatural	O
or	O
too	O
distant	O
from	O
the	O
original	O
colors	O
.	O
</s>
<s>
A	O
way	O
of	O
combating	O
this	O
issue	O
is	O
to	O
use	O
color	B-Algorithm
normalization	I-Algorithm
in	O
combination	O
with	O
thresholding	B-Algorithm
to	O
correctly	O
and	O
consistently	O
segment	O
a	O
colored	O
image	O
.	O
</s>
<s>
There	O
is	O
a	O
vast	O
array	O
of	O
different	O
transformations	O
and	O
algorithms	O
for	O
achieving	O
color	B-Algorithm
normalization	I-Algorithm
and	O
a	O
limited	O
list	O
is	O
presented	O
here	O
.	O
</s>
<s>
As	O
mentioned	O
above	O
,	O
grey	O
world	O
color	B-Algorithm
normalization	I-Algorithm
is	O
invariant	O
to	O
illuminated	O
color	O
variations	O
α	O
,	O
β	O
and	O
γ	O
,	O
however	O
it	O
has	O
one	O
important	O
problem	O
:	O
it	O
does	O
not	O
account	O
for	O
all	O
variations	O
of	O
illumination	O
intensity	O
and	O
it	O
is	O
not	O
dynamic	O
;	O
when	O
new	O
objects	O
appear	O
in	O
the	O
scene	O
it	O
fails	O
.	O
</s>
<s>
Histogram	B-Algorithm
equalization	I-Algorithm
is	O
a	O
non-linear	O
transform	O
which	O
maintains	O
pixel	O
rank	O
and	O
is	O
capable	O
of	O
normalizing	O
for	O
any	O
monotonically	O
increasing	O
color	O
transform	O
function	O
.	O
</s>
<s>
The	O
results	O
of	O
histogram	B-Algorithm
equalization	I-Algorithm
tend	O
to	O
have	O
an	O
exaggerated	O
blue	O
channel	O
and	O
look	O
unnatural	O
,	O
due	O
to	O
the	O
fact	O
that	O
in	O
most	O
images	O
the	O
distribution	O
of	O
the	O
pixel	O
values	O
is	O
usually	O
more	O
similar	O
to	O
a	O
Gaussian	O
distribution	O
,	O
rather	O
than	O
uniform	O
.	O
</s>
<s>
Histogram	B-Algorithm
specification	I-Algorithm
transforms	O
the	O
red	O
,	O
green	O
and	O
blue	O
histograms	O
to	O
match	O
the	O
shapes	O
of	O
three	O
specific	O
histograms	O
,	O
rather	O
than	O
simply	O
equalizing	O
them	O
.	O
</s>
<s>
Histogram	B-Algorithm
specification	I-Algorithm
has	O
the	O
advantage	O
of	O
producing	O
more	O
realistic	O
looking	O
images	O
,	O
as	O
it	O
does	O
not	O
exaggerate	O
the	O
blue	O
channel	O
like	O
histogram	B-Algorithm
equalization	I-Algorithm
.	O
</s>
<s>
The	O
comprehensive	O
color	B-Algorithm
normalization	I-Algorithm
is	O
shown	O
to	O
increase	O
localization	O
and	O
object	O
classification	O
results	O
in	O
combination	O
with	O
color	O
indexing	O
.	O
</s>
<s>
Comprehensive	O
color	B-Algorithm
normalization	I-Algorithm
,	O
just	O
like	O
the	O
histogram	B-Algorithm
equalization	I-Algorithm
method	O
previously	O
mentioned	O
,	O
produces	O
results	O
that	O
may	O
look	O
less	O
natural	O
due	O
to	O
the	O
reduction	O
in	O
the	O
number	O
of	O
color	O
values	O
.	O
</s>
