EE 4780

Edge Detection
Detection of Discontinuities
 Matched Filter Example
>> a=[0 0 0 0 1 2 3 0 0 0 0 2 2 2 0 0 0 0 1 2 -2 -1 0 0 0 0];
>> figure; plot(a);
>> h1 = [-1 -2 2 1]/10;
>> b1 = conv(a,h1); figure; plot(b1);

Bahadir K. Gunturk 2
Detection of Discontinuities
 Point Detection Example:
 Apply a high-pass filter.
 A point is detected if the response is larger than a positive
threshold.
| R |> T

Threshold

 The idea is that the gray level of an isolated point will be quite
different from the gray level of its neighbors.

Bahadir K. Gunturk 3
Detection of Discontinuities
 Point Detection

Detected point

Bahadir K. Gunturk 4
Detection of Discontinuities
 Line Detection Example:

R1 R2 R3 R4

Bahadir K. Gunturk 5
Detection of Discontinuities
 Line Detection Example:

Bahadir K. Gunturk 6
Detection of Discontinuities
 Edge Detection:
 An edge is the boundary between two regions with relatively
distinct gray levels.
 Edge detection is by far the most common approach for
detecting meaningful discontinuities in gray level. The reason
is that isolated points and thin lines are not frequent
occurrences in most practical applications.
 The idea underlying most edge detection techniques is the
computation of a local derivative operator.

Bahadir K. Gunturk 7
Origin of Edges

surface normal discontinuity

depth discontinuity

surface color discontinuity

illumination discontinuity

 Edges are caused by a variety of factors

Bahadir K. Gunturk 8
Profiles of image intensity
edges

Bahadir K. Gunturk 9
Image gradient
 The gradient of an image:

 The gradient points in the direction of most rapid change in intensity

 The gradient direction is given by:

 The edge strength is given by the gradient magnitude

Bahadir K. Gunturk 10
The discrete gradient
 How can we differentiate a digital image f[x,y]?
 Option 1: reconstruct a continuous image, then take gradient
 Option 2: take discrete derivative (finite difference)

Bahadir K. Gunturk 11
Effects of noise
 Consider a single row or column of the image
 Plotting intensity as a function of position gives a signal

Bahadir K. Gunturk 12
Solution: smooth first

Bahadir K. Gunturk Look for peaks in 13

 Derivative theorem of
convolution
 This saves us one operation:

Bahadir K. Gunturk 14
Laplacian of Gaussian
 Consider

Laplacian of Gaussian
operator

Bahadir K. Gunturk Zero-crossings of bottom graph

15
2D edge detection filters

Laplacian of Gaussian

Gaussian derivative of Gaussian

 is the Laplacian operator:

Bahadir K. Gunturk 16
Edge Detection
Possible filters to find gradients along vertical and horizontal directions:

Averaging provides noise

suppression

This gives more importance to the

center point.

Bahadir K. Gunturk 17
Edge Detection

Bahadir K. Gunturk 18
Edge Detection

Bahadir K. Gunturk 19
Edge Detection
 The Laplacian of an image f(x,y) is a second-order
derivative defined as
�2 f �2 f
�f = 2 + 2
2

�x �y

Digital approximations:

Bahadir K. Gunturk 20
Edge Detection

One simple method to find zero-

crossings is black/white thresholding:
1. Set all positive values to white
2. Set all negative values to black
3. Determine the black/white
transitions.

Compare (b) and (g):

•Edges in the zero-crossings image is
thinner than the gradient edges.
•Edges determined by zero-crossings
have formed many closed loops.

Bahadir K. Gunturk 21
Edge Detection
 The Laplacian of a Gaussian filter

A digital approximation:

0 0 1 0 0

0 1 2 1 0

1 2 -16 2 1

0 1 2 1 0

0 0 1 0 0

Bahadir K. Gunturk 22
The Canny edge detector

 original image (Lena)

Bahadir K. Gunturk 23
The Canny edge detector

 norm of the gradient

Bahadir K. Gunturk 24
The Canny edge detector

 thresholding
Bahadir K. Gunturk 25
The Canny edge detector

 thinning
 (non-maximum suppression)
Bahadir K. Gunturk 26
Edge detection by
subtraction

Bahadir K. Gunturk
original 27
Edge detection by
subtraction

Bahadir K. Gunturk
smoothed (5x5 Gaussian) 28
Edge detection by
subtraction

Why does
this work?

smoothed – original
Bahadir K. Gunturk 29
Gaussian - image filter

Gaussian delta function

Bahadir K. Gunturk 30
Laplacian of Gaussian