Scribd
Upload
110 views

Laplacian in Computer Vision

The document discusses the Laplacian operator and zero-crossing operators. The Laplacian is a 2D generalization of the second derivative and measures how a function deviates from being planar near a point. It can be approximated numerically using 3x3 masks on digital images. Taking the Laplacian of a Gaussian smoothed image is equivalent to convolving with a Laplacian of Gaussian (LOG) filter. A pixel is declared to have a zero crossing if its value is less than a threshold and a neighbor's value is greater than the threshold or vice versa.

Uploaded by

R Ravi Teja
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
110 views

Laplacian in Computer Vision

The document discusses the Laplacian operator and zero-crossing operators. The Laplacian is a 2D generalization of the second derivative and measures how a function deviates from being planar near a point. It can be approximated numerically using 3x3 masks on digital images. Taking the Laplacian of a Gaussian smoothed image is equivalent to convolving with a Laplacian of Gaussian (LOG) filter. A pixel is declared to have a zero crossing if its value is less than a threshold and a neighbor's value is greater than the threshold or vice versa.

Uploaded by

R Ravi Teja
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 5

Laplacian

The non-maxima suppression can be incorporated into and made integral part
of the edge operators. Such operators are called zero-crossing operators
The place where first derivative of step is maximum is exactly where second
derivative of the step has a zero crossing.
The isometric generalization of second step to two dimension is the Laplacian.
The laplacian of a function I(r,c) is defined by:
2
2
2
2
2
2
2
2
2
) (
c
I
r
I
I
c r
I
c
c
+
c
c
=
c
c
+
c
c
= V

two common 3X3 masks to calculate digital Laplacian are
First type second type


the 3X3 neighborhood values of an image function are



(-1,-1) (-1, 0) (-1, 1)
( 0,-1) ( 0, 0) ( 0, 1)
( 1,-1) ( 1, 0) ( 1, 1)

2
6 5
2
4 3 2 1
) 1 ( ) 1 )( 1 ( ) 1 ( ) 1 ( ) 1 ( ) 1 , 1 ( + + + + + = k k k k k k I
2
6 5
2
4 3 2 1
) 0 ( ) 0 )( 1 ( ) 1 ( ) 0 ( ) 1 ( ) 0 , 1 ( k k k k k k I + + + + + =
2
6 5
2
4 3 2 1
) 1 ( ) 1 )( 1 ( ) 1 ( ) 1 ( ) 1 ( ) 1 , 1 ( + + + + + = k k k k k k I
2
6 5
2
4 3 2 1
) 0 ( ) 0 )( 1 ( ) 1 ( ) 0 ( ) 1 ( ) 0 , 1 ( k k k k k k I + + + + + =
2
6 5
2
4 3 2 1
) 1 ( ) 1 )( 1 ( ) 1 ( ) 1 ( ) 1 ( ) 1 , 1 ( k k k k k k I + + + + + =


the 3*3 values of I


























For each value of mask produces a correct value of Laplacian of I, which is
2k4+2k6.

2
2
2
2
2
2
2
2
2
) (
c
I
r
I
I
c r
I
c
c
+
c
c
=
c
c
+
c
c
= V
2
6 5
2
4 3 2 1
) , ( c k rc k r k c k r k k c r I + + + + + =
5 4 2
2 ck rk k
r
I
+ + =
c
c
4
2
2
2k
r
I
=
c
c
6 5 3
2ck rk k
c
I
+ + =
c
c
6
2
2
2k
c
I
=
c
c
6 4
2
2
2
2
2
2
2
2
2
2 2 ) ( k k
c
I
r
I
I
c r
I + =
c
c
+
c
c
=
c
c
+
c
c
= V






















3X3 mask computing a digital Laplacian
1 2 ), 4 4 ( = + + = b a b a e





) (k )- k k (k ) k -k (k ) k k (k ) k -k (k
1 4 2 1 4 2 1 6 3 1 6 3 1
4 1 1 1 1 + + + + + + + + +
6 4
2 2 k k + =
] 1 1 1
1 8 1
1 1 1 [
3
1
6 5 4 3 2 1 6 5 4 3 2 1 4 2 1
6 3 1 1 6 3 1
6 5 4 3 2 1 4 2 1 6 5 4 3 2 1
) k k k k k (k ) k -k k -k k (k ) k k (k
) k k (k ) (k )- k -k (k
) k -k k k -k (k ) k -k (k ) k k k -k -k (k
+ + + + + + + + + + + + +
+ + + + +
+ + + + + + + + +
| |
6 4 6 4
2 2 6 6
3
1
k k k k + = + =
`













3X3 mask for computing minimum-variance digital Laplacian











) 2 4 (
) 0 (
) 2 4 (
) 0 (
) 0 (
) 4 4 (
6
5
4
3
2
1
b a k
k
b a k
k
k
e b a k
+
+
+ +
+
+
+ + +
6 4
2 2 k k + =
{
) 4 4 (
1 2
b a - e
b a
+ =
= +
First type
First type
smoothing an image by convolving with Gaussian kernel and then taking its
laplacian by convolving gives laplacian kernel which is exactly same as taking
laplacian of the Gaussian (LOG) kernel and convolving the iamge with it.
The resultant operator is called laplacian of the Gaussian zero corssing
operator




A pixel is declared to have a zero crossing if it is less than t and one of its
eight neighbors is greater than t or if it is greater than t and one of its eight
neighbors is less than t for some fixed threshold t

) (
2
1
2
2
2 2
2
1
:
o
to
c r
e Gaussian
+

2
2
2
2
:
c
I
r
I
Laplacian
c
c
+
c
c
2
2
2
2
) , (
c
G
r
G
c r LOG
c
c
+
c
c
=

You might also like