<s>
Video	B-General_Concept
copy	I-General_Concept
detection	I-General_Concept
is	O
the	O
process	O
of	O
detecting	O
illegally	O
copied	O
videos	O
by	O
analyzing	O
them	O
and	O
comparing	O
them	O
to	O
original	O
content	O
.	O
</s>
<s>
produced	O
a	O
video	B-General_Concept
copy	I-General_Concept
detection	I-General_Concept
theory	O
based	O
on	O
the	O
length	O
of	O
the	O
film	O
;	O
however	O
,	O
it	O
worked	O
only	O
for	O
whole	O
films	O
without	O
modifications	O
.	O
</s>
<s>
introduced	O
the	O
concept	O
of	O
a	O
fingerprint	B-Algorithm
,	O
or	O
hash	B-Error_Name
function	I-Error_Name
,	O
that	O
creates	O
a	O
unique	O
signature	O
of	O
the	O
video	O
based	O
on	O
its	O
contents	O
.	O
</s>
<s>
This	O
fingerprint	B-Algorithm
is	O
based	O
on	O
the	O
length	O
of	O
the	O
video	O
and	O
the	O
brightness	O
,	O
as	O
determined	O
by	O
splitting	O
it	O
into	O
a	O
grid	O
.	O
</s>
<s>
The	O
fingerprint	B-Algorithm
cannot	O
be	O
used	O
to	O
recreate	O
the	O
original	O
video	O
because	O
it	O
describes	O
only	O
certain	O
features	O
of	O
its	O
respective	O
video	O
.	O
</s>
<s>
presented	O
two	O
robust	O
algorithms	O
based	O
on	O
discrete	B-General_Concept
cosine	I-General_Concept
transform	I-General_Concept
.	O
</s>
<s>
This	O
algorithm	O
,	O
however	O
,	O
returns	O
inconsistent	O
results	O
in	O
cases	O
in	O
which	O
a	O
logo	B-Device
is	O
added	O
to	O
the	O
video	O
,	O
because	O
the	O
insertion	O
of	O
the	O
logo	B-Device
's	O
color	O
elements	O
adds	O
false	O
information	O
that	O
can	O
confuse	O
the	O
system	O
.	O
</s>
<s>
Watermarks	B-General_Concept
are	O
used	O
to	O
introduce	O
an	O
invisible	O
signal	O
into	O
a	O
video	O
to	O
ease	O
the	O
detection	O
of	O
illegal	O
copies	O
.	O
</s>
<s>
Placing	O
a	O
watermark	B-General_Concept
on	O
a	O
video	O
such	O
that	O
it	O
is	O
easily	O
seen	O
by	O
an	O
audience	O
allows	O
the	O
content	O
creator	O
to	O
detect	O
easily	O
whether	O
the	O
image	O
has	O
been	O
copied	O
.	O
</s>
<s>
The	O
limitation	O
of	O
watermarks	B-General_Concept
is	O
that	O
if	O
the	O
original	O
image	O
is	O
not	O
watermarked	O
,	O
then	O
it	O
is	O
not	O
possible	O
to	O
know	O
whether	O
other	O
images	O
are	O
copies	O
.	O
</s>
<s>
Various	O
video	B-General_Concept
copy	I-General_Concept
detection	I-General_Concept
algorithms	O
exist	O
that	O
use	O
features	O
of	O
the	O
video	O
's	O
content	O
to	O
assign	O
the	O
video	O
a	O
unique	O
videohash	B-Algorithm
.	O
</s>
<s>
The	O
fingerprint	B-Algorithm
can	O
be	O
compared	O
with	O
other	O
videohashes	O
in	O
a	O
database	O
.	O
</s>
<s>
Videos	O
with	O
very	O
minimal	O
changes	O
in	O
frames	O
with	O
respect	O
to	O
time	O
are	O
more	O
vulnerable	O
to	O
hash	B-Error_Name
collision	O
.	O
</s>
<s>
The	O
following	O
are	O
some	O
algorithms	O
and	O
techniques	O
proposed	O
for	O
video	B-General_Concept
copy	I-General_Concept
detection	I-General_Concept
.	O
</s>
<s>
It	O
generates	O
a	O
signature	O
of	O
a	O
whole	O
video	O
,	O
and	O
can	O
easily	O
handle	O
changes	O
in	O
format	O
or	O
added	O
watermarks	B-General_Concept
,	O
but	O
is	O
less	O
tolerant	O
of	O
cropping	O
or	O
clipping	O
.	O
</s>
<s>
There	O
exist	O
algorithms	O
for	O
video	B-General_Concept
copy	I-General_Concept
detection	I-General_Concept
that	O
are	O
in	O
use	O
today	O
.	O
</s>
<s>
In	O
2007	O
,	O
there	O
was	O
an	O
evaluation	O
showcase	O
known	O
as	O
the	O
Multimedia	O
Understanding	O
Through	O
Semantics	O
,	O
Computation	O
and	O
Learning	O
(	O
MUSCLE	O
)	O
,	O
which	O
tested	O
video	B-General_Concept
copy	I-General_Concept
detection	I-General_Concept
algorithms	O
on	O
various	O
video	O
samples	O
ranging	O
from	O
home	O
video	O
recordings	O
to	O
TV	O
show	O
segments	O
ranging	O
from	O
one	O
minute	O
to	O
one	O
hour	O
in	O
length	O
.	O
</s>
