[go: up one dir, main page]
Scribd
Upload
73 views30 pages

Lec 2 Blob - Features

The document discusses the Laplacian of Gaussian (LoG) operator for blob detection. [1] LoG is a circularly symmetric operator that finds maxima and minima of the second spatial derivatives of an image smoothed by a Gaussian filter at different scales. [2] The characteristic scale for peak LoG response to a blob is when the Gaussian scale matches the size of the blob. [3] An efficient approximation of LoG is the difference of Gaussians (DoG) operator, which computes the difference between images smoothed with Gaussian kernels of different standard deviations.

Uploaded by

ahmed sayed
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
73 views30 pages

Lec 2 Blob - Features

The document discusses the Laplacian of Gaussian (LoG) operator for blob detection. [1] LoG is a circularly symmetric operator that finds maxima and minima of the second spatial derivatives of an image smoothed by a Gaussian filter at different scales. [2] The characteristic scale for peak LoG response to a blob is when the Gaussian scale matches the size of the blob. [3] An efficient approximation of LoG is the difference of Gaussians (DoG) operator, which computes the difference between images smoothed with Gaussian kernels of different standard deviations.

Uploaded by

ahmed sayed
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 30

Laplacian of Gaussian

Blob Features

Lecture 2
Automatic scale selection

2
The Laplacian of Gaussian (LoG)

∂2 g ∂2 g
∇g =
2
+
∂x2 ∂y2
3
Edges and blobs

4
Edges and blobs

5
Edges and blobs

• Edge: Ripple
• Blob: Superposition of two ripples

• The magnitude of the Laplacian response is maximum at the centre of the blob
provided the scale of the Laplacian matches the scale of the blob
6
Characteristic scale

• We define the characteristic scale as the scale that produces peak of Laplacian response

8
Characteristic scale

9
Scale selection

• At what scale does the Laplacian achieve a maximum response to a binary


circle of radius r?

10
Scale selection

• To get maximum response, the zeros of the Laplacian have to be aligned with
the circle
• The Laplacian is given by (up to scale):
(x2 + y2 − 2σ2 ) e−( x2 + y2 ) / 2σ2
• Therefore, the maximum response occurs at σ= r / 2.

11
Laplacian of Gaussian

• Circularly symmetric operator for blob detection in 2D

• Find maxima and minima of LoG operator in space and scale 12


LoG blob detector

• Example

14
LoG blob detector

• Example: Squared Log response

15
LoG blob detector

• Example: Squared Log response

16
LoG blob detector

• Example: Squared Log response

17
LoG blob detector

• Example: Squared Log response

18
LoG blob detector

• Example: Squared Log response

19
LoG blob detector

• Example: Squared Log response

20
LoG blob detector

• Example: Squared Log response

21
LoG blob detector

• Example: Squared Log response

22
LoG blob detector

• Example: Squared Log response

23
LoG blob detector

• Example: Squared Log response

24
LoG blob detector

• Example: Detected blobs

25
26
Efficient implementation

• Approximating the Laplacian with a difference of Gaussians (DoG)

L =σ2 (Gxx (x, y,σ) + Gyy (x, y,σ))


(Laplacian)

DoG = G(x, y, kσ) − G(x, y,σ)


(Difference of Gaussians)

27
Efficient implementation

28
Difference of Gaussians (DoG)
How does this work?
• Taking the difference between a blurred image
and original
Difference of Gaussians (DoG)

-
29

You might also like