Mathematical Morphology

A Geometric Approach to Image Processing and Analysis

John Goutsias
Department of Electrical and Computer Engineering
Image Analysis and Communications Laboratory
The Johns Hopkins University
Baltimore, MD 21218
Question

What is Mathematical Morphology ?

2
A Commercial Answer

 Mathematical Morphology is FAST !!

 Mathematical Morphology is CHEAP !!

3
An (imprecise) Mathematical Answer

A mathematical tool for investigating geometric

structure in binary and grayscale images.

4
Shape Processing and Analysis

 Visual perception requires transformation of images

so as to make explicit particular shape information.

 Goal: Distinguish meaningful shape information

from irrelevant one.

 The vast majority of shape processing and analysis

techniques are based on designing a shape operator
which satisfies desirable properties.

5
Example

 Image analysis consists of obtaining

measurements characteristic to images
under consideration.
Grayscale Images
 Geometric measurements (e.g., object
location, orientation, area, length of
perimeter)

Binary Images

6
 Morphological Shape Operators

 Objects are opaque and shape information is not additive !!

 Shapes are usually combined by means of:

 Set Union (overlapping objects):

X1 X2 X1  X2

 Set Intersection (occluded objects):

X1 X2 X2 \ X1  X1c  X2

7
Morphological Shape Operators

 Shape operators should distribute over set-unions and

set-intersections (a type of “linearity”) !

Morphological
Dilation
 ( X1  X2 ) =  ( X1 )  ( X2 )

Morphological
Erosion
 ( X1  X2 ) =  ( X1 )  ( X2 )

8
Morphological Operators

 Erosions and dilations are the most elementary

operators of mathematical morphology.

 More complicated morphological operators can

be designed by means of combining erosions and
dilations.

9
Question

What is Mathematical Morphology ?

10
A (precise) Mathematical Answer

Algebra Topology
Complete Lattices Hit-or-Miss

Mathematical
Morphology

Geometry
Operators
Convexity - Connectivity
Erosions-Dilations
Distance

Applications
Image Processing
and Analysis

AAmathematical
mathematicaltool
toolthat
thatstudies
studiesoperators
operatorson
on
complete
completelattices
lattices
11
 Some History

 George Matheron (1975) Random Sets and Integral Geometry,

John Wiley.

 Jean Serra (1982) Image Analysis and Mathematical Morphology,

Academic Press.

 Petros Maragos (1985) A Unified Theory of Translations-Invariant

Systems with Applications to Morphological Analysis and Coding of
Images, Doctoral Thesis, Georgia Tech.

12
Translation Invariant Operators

Xh
h
( Xh ) = [ ( X )]h
X

13
Morphological Erosion

“LINEARITY” TRANSLATION INVARIANCE

 ( X1  X2 ) =  ( X1 )  ( X2 )  ( Xh ) = [ ( X )]h

 ( X ) = X  B  { h | Bh  X }

14
Morphological Erosion

 ( X ) = X  B  { h | Bh  X }

Bh
X

X B

15
Morphological Erosion

Structuring
Element

Pablo Picasso, Pass with the Cape, 1960

16
Morphological Dilation

“LINEARITY” TRANSLATION INVARIANCE

 ( X1  X2 ) =  ( X1 )  ( X2 )  ( Xh ) = [ ( X )]h

 ( X ) = X  B  {h | Bh X  }

17
Morphological Dilation

 ( X ) = X  B  {h | Bh  X  }

Bh

XB
X

18
Morphological Dilation

Structuring
Element

Pablo Picasso, Pass with the Cape, 1960

19
Morphological Opening

X  B = ( X  B)  B
 {Bh | Bh  X}
X  B = ( X  B)  B X

X B

Bh
20
Morphological Opening

Structuring
Element

Pablo Picasso, Pass with the Cape, 1960

21
Morphological Opening

 Is a smoothing filter !

 Amount and type of smoothing is determined by

the shape and size of the structuring element.

 Approximates a shape from below, since

X B X

22
Filtering Example

ORIGINAL DEGRADED

Henri Matisse, Woman with Amphora

and Pomegranates, 1952

FILTERED 23
An Important Result

Increasing
Operator
X1  X2  ( X1 )  ( X2 )

+
Translation
Invariant ( Xh ) = [( X )]h
Operator

( X ) =  X  B   X  B
B B
!!
24
Main Idea

 Examine the geometrical structure of an image by

matching it with small patterns at various locations.

 By varying the size and shape of the matching patterns,

called structuring elements, one can extract useful
information about the shape of the different parts of the
image and their interrelations.

 Results in image operators which are well suited for

the analysis of the geometrical and topological
structure of an image.

25
Question

What about gray-scale images ?

26
 Grayscale Erosion

TRANSLATION INVARIANCE

“LINEARITY”  ( F ( x  h)) = [ ( F )]( x  h)

 ( F1  F2 ) =  ( F1 )   ( F2 )  ( F ( x )  v ) = [ ( F )]( x )  v

MINIMUM

[ ( F )]( x ) = F  B ( x )   [F(h)  B (h  x)]

h

27
 Grayscale Dilation

TRANSLATION INVARIANCE

“LINEARITY”  ( F( x  h)) = [ ( F )]( x  h)

 ( F1  F2 ) =  ( F1 )   ( F2 )  ( F( x )  v) = [ ( F )]( x )  v

MAXIMUM

[ ( F )]( x ) = F  B( x )   [ F(h)  B( x  h)]

h

28
Remark

R
B( x )  S
0, for x  B
T, otherwise

Flat
Erosion
F  B( x)   F(h)
h Bx

Flat
Dilation
F  B( x)   F (h)
h Bx

29
Grayscale Morphology

ORIGINAL EROSION

DILATION OPENING
30
Grayscale Opening

Structuring
Element

31
Question

Henri Matisse, Woman with Amphora

and Pomegranates, 1952

Can
Canweweautomatically
automaticallyextract
extractthe
thelargest
largest
connected
connectedcomponent
component(the(thewoman’s
woman’s
body)
body)in
inthis
thisimage
image??

32
 Answer
ORIGINAL

(MARKER  B) ORIGINAL

This is a morphological operator that filters out

EROSION connected image components of a certain

MARKER
size and shape
MARKER
MARKER
MARKER

CONNECTED OPERATORS !!

33
 An Application - Target Detection

DATA
Targets
MORPHOLOGICAL RECONSTRUCTION

OPENING

MARKER

34
 An Application: Target Detection

DATA

MORPHOLOGICAL RECONSTRUCTION

CLOSING

MARKER

35
 An Application: Target Detection
DATA

FINAL RESULT

THRESHOLDING

Incorrectly detected target

Correctly detected targets

36
The End
37