<s>
The	O
Harris	B-General_Concept
corner	I-General_Concept
detector	I-General_Concept
is	O
a	O
corner	B-Algorithm
detection	I-Algorithm
operator	O
that	O
is	O
commonly	O
used	O
in	O
computer	B-Application
vision	I-Application
algorithms	O
to	O
extract	O
corners	O
and	O
infer	O
features	B-Algorithm
of	O
an	O
image	O
.	O
</s>
<s>
It	O
was	O
first	O
introduced	O
by	O
Chris	O
Harris	O
and	O
Mike	O
Stephens	O
in	O
1988	O
upon	O
the	O
improvement	O
of	O
Moravec	O
's	O
corner	B-Algorithm
detector	I-Algorithm
.	O
</s>
<s>
Compared	O
to	O
its	O
predecessor	O
,	O
Harris	O
 '	O
corner	B-Algorithm
detector	I-Algorithm
takes	O
the	O
differential	O
of	O
the	O
corner	O
score	O
into	O
account	O
with	O
reference	O
to	O
direction	O
directly	O
,	O
instead	O
of	O
using	O
shifting	O
patches	O
for	O
every	O
45	O
degree	O
angles	O
,	O
and	O
has	O
been	O
proved	O
to	O
be	O
more	O
accurate	O
in	O
distinguishing	O
between	O
edges	O
and	O
corners	O
.	O
</s>
<s>
Corners	O
are	O
the	O
important	O
features	B-Algorithm
in	O
the	O
image	O
,	O
and	O
they	O
are	O
generally	O
termed	O
as	O
interest	O
points	O
which	O
are	O
invariant	O
to	O
translation	O
,	O
rotation	O
and	O
illumination	O
.	O
</s>
<s>
Although	O
corners	O
are	O
only	O
a	O
small	O
percentage	O
of	O
the	O
image	O
,	O
they	O
contain	O
the	O
most	O
important	O
features	B-Algorithm
in	O
restoring	O
image	O
information	O
,	O
and	O
they	O
can	O
be	O
used	O
to	O
minimize	O
the	O
amount	O
of	O
processed	O
data	O
for	O
motion	O
tracking	O
,	O
image	B-Algorithm
stitching	I-Algorithm
,	O
building	O
2D	O
mosaics	O
,	O
stereo	B-Algorithm
vision	I-Algorithm
,	O
image	O
representation	O
and	O
other	O
related	O
computer	B-Application
vision	I-Application
areas	O
.	O
</s>
<s>
In	O
order	O
to	O
capture	O
the	O
corners	O
from	O
the	O
image	O
,	O
researchers	O
have	O
proposed	O
many	O
different	O
corner	B-Algorithm
detectors	I-Algorithm
including	O
the	O
Kanade-Lucas-Tomasi	O
(	O
KLT	O
)	O
operator	O
and	O
the	O
Harris	O
operator	O
which	O
are	O
most	O
simple	O
,	O
efficient	O
and	O
reliable	O
for	O
use	O
in	O
corner	B-Algorithm
detection	I-Algorithm
.	O
</s>
<s>
Compared	O
to	O
the	O
Kanade-Lucas-Tomasi	O
corner	B-Algorithm
detector	I-Algorithm
,	O
the	O
Harris	B-General_Concept
corner	I-General_Concept
detector	I-General_Concept
provides	O
good	O
repeatability	O
under	O
changing	O
illumination	O
and	O
rotation	O
,	O
and	O
therefore	O
,	O
it	O
is	O
more	O
often	O
used	O
in	O
stereo	O
matching	O
and	O
image	O
database	O
retrieval	O
.	O
</s>
<s>
Although	O
there	O
still	O
exists	O
drawbacks	O
and	O
limitations	O
,	O
the	O
Harris	B-General_Concept
corner	I-General_Concept
detector	I-General_Concept
is	O
still	O
an	O
important	O
and	O
fundamental	O
technique	O
for	O
many	O
computer	B-Application
vision	I-Application
applications	O
.	O
</s>
<s>
where	O
M	O
is	O
the	O
structure	B-Algorithm
tensor	I-Algorithm
,	O
</s>
<s>
Commonly	O
,	O
Harris	B-General_Concept
corner	I-General_Concept
detector	I-General_Concept
algorithm	O
can	O
be	O
divided	O
into	O
five	O
steps	O
.	O
</s>
<s>
If	O
we	O
use	O
Harris	B-General_Concept
corner	I-General_Concept
detector	I-General_Concept
in	O
a	O
color	O
image	O
,	O
the	O
first	O
step	O
is	O
to	O
convert	O
it	O
into	O
a	O
grayscale	O
image	O
,	O
which	O
will	O
enhance	O
the	O
processing	O
speed	O
.	O
</s>
<s>
With	O
,	O
,	O
we	O
can	O
construct	O
the	O
structure	B-Algorithm
tensor	I-Algorithm
.	O
</s>
<s>
In	O
this	O
step	O
,	O
we	O
compute	O
the	O
smallest	O
eigenvalue	O
of	O
the	O
structure	B-Algorithm
tensor	I-Algorithm
using	O
that	O
approximation	O
:	O
</s>
