<s>
In	O
digital	B-Algorithm
image	I-Algorithm
and	O
video	O
processing	O
,	O
a	O
color	B-Algorithm
layout	I-Algorithm
descriptor	I-Algorithm
(	O
CLD	O
)	O
is	O
designed	O
to	O
capture	O
the	O
spatial	O
distribution	O
of	O
color	O
in	O
an	O
image	O
.	O
</s>
<s>
The	O
feature	O
extraction	O
process	O
consists	O
of	O
two	O
parts	O
:	O
grid	O
based	O
representative	O
color	O
selection	O
and	O
discrete	B-General_Concept
cosine	I-General_Concept
transform	I-General_Concept
with	O
quantization	O
.	O
</s>
<s>
This	O
standard	O
proposes	O
different	O
methods	O
to	O
obtain	O
these	O
descriptors	B-General_Concept
,	O
and	O
one	O
tool	O
defined	O
to	O
describe	O
the	O
color	O
is	O
the	O
CLD	O
,	O
that	O
permits	O
describing	O
the	O
color	O
relation	O
between	O
sequences	O
or	O
group	O
of	O
images	O
.	O
</s>
<s>
Representation	O
is	O
based	O
on	O
coefficients	O
of	O
the	O
discrete	B-General_Concept
cosine	I-General_Concept
transform	I-General_Concept
(	O
DCT	B-General_Concept
)	O
.	O
</s>
<s>
The	O
CLD	O
is	O
a	O
very	O
compact	O
and	O
resolution-invariant	O
representation	O
of	O
color	O
for	O
high-speed	O
image	B-General_Concept
retrieval	I-General_Concept
and	O
it	O
has	O
been	O
designed	O
to	O
efficiently	O
represent	O
the	O
spatial	O
distribution	O
of	O
colors	O
.	O
</s>
<s>
This	O
descriptor	O
is	O
obtained	O
by	O
applying	O
the	O
DCT	B-General_Concept
transformation	O
on	O
a	O
2-D	O
array	O
of	O
local	O
representative	O
colors	O
in	O
Y	O
or	O
Cb	O
or	O
Cr	O
color	B-Algorithm
space	I-Algorithm
.	O
</s>
<s>
The	O
standard	O
MPEG-7	O
recommends	O
using	O
the	O
YCbCr	O
color	B-Algorithm
space	I-Algorithm
for	O
the	O
CLD	O
.	O
</s>
<s>
In	O
the	O
image	O
partitioning	O
stage	O
,	O
the	O
input	O
picture	O
(	O
on	O
RGB	O
color	B-Algorithm
space	I-Algorithm
)	O
is	O
divided	O
into	O
64	O
blocks	O
to	O
guarantee	O
the	O
invariance	O
to	O
resolution	O
or	O
scale	O
.	O
</s>
<s>
Once	O
the	O
tiny	O
image	O
icon	O
is	O
obtained	O
,	O
the	O
color	B-Algorithm
space	I-Algorithm
conversion	I-Algorithm
between	O
RGB	O
and	O
YCbCr	O
is	O
applied	O
.	O
</s>
<s>
In	O
the	O
fourth	O
stage	O
,	O
the	O
luminance	O
(	O
Y	O
)	O
and	O
the	O
blue	O
and	O
red	O
chrominance	O
(	O
Cb	O
and	O
Cr	O
)	O
are	O
transformed	O
by	O
8x8	O
DCT	B-General_Concept
,	O
so	O
three	O
sets	O
of	O
64	O
DCT	B-General_Concept
coefficients	O
are	O
obtained	O
.	O
</s>
<s>
To	O
calculate	O
the	O
DCT	B-General_Concept
in	O
a	O
2D	O
array	O
,	O
the	O
formulas	O
below	O
are	O
used	O
.	O
</s>
<s>
A	O
zigzag	O
scanning	O
is	O
performed	O
with	O
these	O
three	O
sets	O
of	O
64	O
DCT	B-General_Concept
coefficients	O
,	O
following	O
the	O
schema	O
presented	O
in	O
the	O
figure	O
.	O
</s>
<s>
In	O
the	O
case	O
of	O
color	O
descriptors	B-General_Concept
the	O
matching	O
process	O
helps	O
to	O
evaluate	O
if	O
two	O
images	O
are	O
similar	O
.	O
</s>
<s>
The	O
distance	O
between	O
the	O
two	O
descriptors	B-General_Concept
can	O
be	O
computed	O
as	O
:	O
</s>
<s>
Therefore	O
,	O
this	O
matching	O
process	O
will	O
let	O
to	O
identify	O
images	O
with	O
similar	O
color	O
descriptors	B-General_Concept
.	O
</s>
<s>
We	O
aim	O
to	O
find	O
images	O
with	O
similar	O
colors	O
,	O
thus	O
,	O
we	O
have	O
to	O
extract	O
the	O
CLD	O
from	O
these	O
images	O
and	O
afterwards	O
compare	O
these	O
descriptors	B-General_Concept
with	O
the	O
matching	O
technique	O
.	O
</s>
