<s>
The	O
principal	B-Algorithm
curvature-based	I-Algorithm
region	I-Algorithm
detector	I-Algorithm
,	O
also	O
called	O
PCBR	O
is	O
a	O
feature	O
detector	O
used	O
in	O
the	O
fields	O
of	O
computer	B-Application
vision	I-Application
and	O
image	B-General_Concept
analysis	I-General_Concept
.	O
</s>
<s>
These	O
detectors	O
include	O
SIFT	B-Algorithm
,	O
Hessian-affine	B-Algorithm
,	O
Harris-Affine	B-Algorithm
and	O
MSER	B-Algorithm
etc	O
.	O
</s>
<s>
From	O
the	O
detection	O
invariance	O
point	O
of	O
view	O
,	O
feature	O
detectors	O
can	O
be	O
divided	O
into	O
fixed	O
scale	O
detectors	O
such	O
as	O
normal	O
Harris	B-General_Concept
corner	I-General_Concept
detector	I-General_Concept
,	O
scale	O
invariant	O
detectors	O
such	O
as	O
SIFT	B-Algorithm
and	O
affine	O
invariant	O
detectors	O
such	O
as	O
Hessian-affine	B-Algorithm
.	O
</s>
<s>
As	O
such	O
,	O
intensity-based	O
interest	O
operators	O
(	O
e.g.	O
,	O
SIFT	B-Algorithm
,	O
Harris-Affine	B-Algorithm
)	O
–	O
and	O
the	O
object	O
recognition	O
systems	O
based	O
on	O
them	O
–	O
often	O
fail	O
to	O
identify	O
discriminative	O
features	O
.	O
</s>
<s>
To	O
make	O
this	O
detector	O
scale	O
invariant	O
and	O
improve	O
the	O
detection	O
robustness	O
,	O
the	O
process	O
of	O
David	O
Lowe	O
's	O
SIFT	B-Algorithm
detector	O
is	O
simulated	O
to	O
detect	O
principal	O
curvilinear	O
structure	O
in	O
scale	B-Algorithm
space	I-Algorithm
.	O
</s>
<s>
Then	O
traditional	O
watershed	B-Algorithm
algorithm	I-Algorithm
is	O
applied	O
on	O
images	O
to	O
acquire	O
regions	O
.	O
</s>
<s>
Similar	O
to	O
the	O
process	O
of	O
selecting	O
stable	O
regions	O
via	O
thresholding	O
in	O
MSER	B-Algorithm
,	O
stable	O
regions	O
are	O
selected	O
across	O
local	O
scale	O
changes	O
.	O
</s>
<s>
These	O
numbers	O
are	O
determined	O
by	O
the	O
analysis	O
of	O
sensitivity	O
of	O
the	O
SIFT	B-Algorithm
descriptor	O
.	O
</s>
