morphoEngAll2 PDF

Mathematical Morphology
Isabelle Bloch
http://www.tsi.enst.fr/bloch
ecommunications
Ecole Nationale Superieure

des Tel
- CNRS UMR 5141 LTCI
Paris - France
Mathematical Morphology p.1/84
A few references
J. Serra, Image Analysis and Mathematical Morphology, Academic Press,

New-York, 1982.
J. Serra (Ed.), Image Analysis and Mathematical Morphology, Part II: Theoretical
Advances, Academic Press, London, 1988.
P. Soille, Morphological Image Analysis, Springer-Verlag, Berlin, 1999.
Shape or spatial relations?
Simplifying and selecting relevant information...
Introduction
Origin: study of porous media

Principle: study of objects (images) based on:
grey levels, colors

neighborhood information
Mathematical bases:
set theory
shape, geometry, topology
topology
geometry
algebra (lattice theory)
probabilities, random closed sets
functions
Main characteristics:
non linear
non invertible
strong properties
associated algorithms
Tools for
filtering
segmentation
measures (distances, granulometry, integral geometry, topology, stochastic
processes...)
texture analysis
shape recognition
scene interpretation
...
Applications in numerous domains
Example
Example
Four fundamental principles
1. Compatibility with translations

2. Compatibility with scaling
3. Local knowledge
4. Continuity (semi-continuity)
Structuring element
shape
size
origin (not necessarily in B)
examples:
Continuous:
Digital:
Binary dilation
Minkowski addition:
X Y = {x + y / x X, y Y }
Binary dilation:
D(X, B)
=
=
= {x + y / x X, y B}
(or = X B)
X B
[
x = {x Rn / Bx X 6= }
B
xX
Binary dilation
Minkowski addition:
X Y = {x + y / x X, y Y }
Binary dilation:
D(X, B)
=
=
= {x + y / x X, y B}
(or = X B)
X B
[
x = {x Rn / Bx X 6= }
B
xX
Properties of dilation:
extensive (X D(X, B)) if O B;

increasing (X Y D(X, B) D(Y, B));
B B 0 D(X, B) D(X, B 0 );
commutes with union, not with intersection:
D(X Y, B) = D(X, B) D(Y, B),
D(X Y, B) D(X, B) D(Y, B);
iterativity property: D[D(X, B), B 0 ] = D(X, B B 0 ).
Example of dilation
Binary erosion
E(X, B)
{x Rn / Bx X}
{x / y B, x + y X} = X B.
Binary erosion
E(X, B)
{x Rn / Bx X}
{x / y B, x + y X} = X B.
Properties of erosion:
duality of erosion and dilation with respect to complementation:

E(X, B) = [D(X C , B)]C
C)
(or E(X, B) = [D(X C , B)]
anti-extensive (E(X, B) X) if O B;
increasing (X Y E(X, B) E(Y, B));
B B 0 E(X, B 0 ) E(X, B);
commutes with intersection, not with union:
E[(X Y ), B] = E(X, B) E(Y, B),
E[(X Y ), B] E(X, B) E(Y, B);
iterativity property: E[E(X, B), B 0 ] = E(X, B B 0 ).

D[E(X, B), B 0 ] E[D(X, B 0 ), B].
Example of erosion
Links with distances
Links with distances
Binary opening
XB = D[E(X, B), B]
(or D[E(X, B), B])
Binary opening
XB = D[E(X, B), B]
(or D[E(X, B), B])
Properties of opening:
anti-extensive (X XB );
increasing (X Y XB YB );
idempotent ((XB )B = XB ).
Morphological filter
B B 0 XB 0 XB ;
(Xn )n0 = (Xn0 )n = Xmax(n,n0 ) .
Example of opening
Binary closing
X B = E[D(X, B), B]
(or E[D(X, B), B])
Binary closing
X B = E[D(X, B), B]
(or E[D(X, B), B])
Properties of closing:
extensive (X X B );
increasing (X Y X B Y B );
idempotent ((X B )B = X B ).
Morphological filter
0
B B0 X B X B ;
(X n )n = (X n )n = X max(n,n ) ;
X B = [(X C )B ]C .
Example of closing
Digital case
choice of the digital grid (both for the image and the structuring element)
translations on the grid
same properties
From sets to functions
subgraph of a function on Rn = subset of Rn+1

cuts of a function = sets
f = {x|f (x) }
D(f , B) = [D(f, B)]
functional equivalents of set operations:
sup /
inf /
Dilation of a function
by a flat structuring element
x Rn , D(f, B)(x) = sup{f (y) / y Bx }
Dilation of a function
by a flat structuring element
x Rn , D(f, B)(x) = sup{f (y) / y Bx }
Properties of functional dilation:
extensivity if O B;
increasingness;
D(f g, B) = D(f, B) D(g, B);
D(f g, B) D(f, B) D(g, B);
iterativity property.
It holds:
D(f , B) = [D(f, B)]
Example of functional dilation
Erosion of a function
x Rn , E(f, B)(x) = inf{f (y) / y Bx }
Erosion of a function
x Rn , E(f, B)(x) = inf{f (y) / y Bx }
Properties of functional erosion:
functional dilation and erosion are dual operators;

anti-extensivity if O B;
increasingness;
E(f g, B) E(f, B) E(g, B);
E(f g, B) = E(f, B) E(g, B);
iterativity property.
Example of functional erosion
Functional opening
fB = D[E(f, B), B]
Functional opening
fB = D[E(f, B), B]
Properties of functional opening:
anti-extensive;
increasing;
idempotent.
morphological filter
Example of functional opening
Functional closing
f B = E[D(f, B), B]
Functional closing
f B = E[D(f, B), B]
Properties of functional closing:
extensive;
increasing;
idempotent.
morphological filter
duality between opening and closing
Example of functional closing
Structuring functions
Dilation:
D(f, g)(x) = sup{f (y) + g(y x)}
y
Erosion:
E(f, g)(x) = inf {f (y) g(y x)}
y
Flat structuring element:
g(x) =
on a compact B
elsewhere
Some applications of erosion and dilation

Contrast enhancement

Contrast enhancement: ES 15, = = 0.2, = = 0.3, = = 0.5

Contrast enhancement: ES 30, = = 0.2, = = 0.3, = = 0.5

Morphological gradient:
DB (x) EB (x)

Ultimate erosion:
EU (X) = n {E(X, Bn ) \ R[E(X, Bn+1 ); E(X, Bn )]}
E(X, Bn ): erosion of X by a structuring element of size n

R[Y ; Z]: connected components of Z having a non-empty intersection with Y
= set of regional maxima of the distance function d(x, X C ).
An application of opening:
top-hat transform
f fB
top-hat transform
f fB
top-hat transform
f fB
Granulometry
X A, > 0, (X) X ( anti-extensive);
> 0, > 0, = = max(,) .
(X, Y ) A2 , > 0, X Y (X) (Y ) ( increasing);

X A, > 0, > 0 (X) (X) ( decreasing with respect
to the parameter);
Granulometry
> 0, > 0, = = max(,) .

to the parameter);
( ) is a granulometry iff is an opening for each and the class of subsets A which
are invariant under is included in the class of subsets which are invariant under for
Granulometry
> 0, > 0, = = max(,) .

to the parameter);
( ) is a granulometry iff is an opening for each and the class of subsets A which
are invariant under is included in the class of subsets which are invariant under for
Vectorial functions (e.g. color images)
Main difficulty: choice of an ordering

component-wise max (or min): no good properties
Dilation
componentwise
maximum
Choice of the structuring element
depends on what one wants suppress / keep

shape
size
Example: opening by disks or segments?

shape
size

shape
size

shape
size
Rq: a union of openings is an opening
Surfacic opening
(f ) =
{Bi (f ), Bi connected and S(Bi ) = }
Surfacic opening
(f ) =

(f ) =
Surfacic opening
Mathematical fundaments of mathematical

morphology
Set theory
hit-or-miss topology (Fells topology)

myopic topology
Hausdorff distance
Lattice theory
structuring element
Topology
relations (, , ...)
adjunctions
algebraic operations
Probability theory
P (A K 6= )
random closed sets
Hit-or-miss topology
topology on closed subsets

generated by F K and FG (K compact and G open):
F K = {F F , F K = }
FG = {F F , F G 6= }
convergence in F : (Fn )nN converges towards F F if:

(
G G, G F 6= , N, n N, G Fn 6=
K K, K F = , N 0 , n N 0 , K Fn =
Hit-or-miss topology
topology on closed subsets

generated by F K and FG (K compact and G open):
F K = {F F , F K = }
FG = {F F , F G 6= }
convergence in F : (Fn )nN converges towards F F if:

(
G G, G F 6= , N, n N, G Fn 6=
K K, K F = , N 0 , n N 0 , K Fn =
Union is continuous from F F in F but intersection is not
semi-continuity
Semi-continuity
f :F
f upper semi-continuous (u.s.c.) if and (n )nN converging towards

:
limf (n ) f ()
f lower semi-continuous (l.s.c.) if:

limf (n ) f ()
lim/lim = / of adherence points

f continuous iff f l.s.c. and u.s.c.
Intersection is u.s.c.
Properties of morphological operations
the dilation of a closed set by a compact set is continuous
(F, K) 7 F K u.s.c.
the dilation of a compact set by a compact set is continuous

(F, K) 7 E(F, K) u.s.c.
(K 0 , K) 7 E(K 0 , K) u.s.c.
(F, K) 7 FK u.s.c.
0 u.s.c.
(K 0 , K) 7 KK
(K 0 , K) 7 K 0K u.s.c.
Myopic topology
generated by:
F
= {K K, K F = , K G 6= }
KG
(F F , G G)
finer than the topology induced on K by the hit-or-miss topology

equivalent on K \ to the topology induced by the Hausdorff distance
(K, K 0 ) = max{ sup d(x, K 0 ), sup d(x0 , K)}
xK
x0 K 0
Rq: (K, K 0 ) = inf{, K D(K 0 , B ), K 0 D(K, B )}
Algebraic framework: complete lattices
Lattice: (T , ) ( ordering) such that (x, y) T , x y and x y

Complete lattice: every family of elements (finite or not) has a smallest upper
bound and a largest lower bound
contains a smallest element 0 and a largest element I:
0=
T =
et I =
T =
Examples of complete lattices:
(P(E), ): complete lattice, Boolean (complemented and distributive):

x, xC , x xC = 0 and x xC = I
x (y z) = (x y) (x z) and x (y z) = (x y) (x z)
(F (Rd ), )
functions of Rn in R for the ordering :
f g x Rn , f (x) g(x)
partitions
Semi-continuity of functions
u.s.c. :
t > f (x), V (x), y V (x), t > f (y)
(V (x) neighborhood of x in Rn )
l.s.c. :
t < f (x), V (x), y V (x), t < f (y)
a function is u.s.c. iff its sub-graph is closed

topology on the space of u.s.c. functions = topology induced by the hit-or-miss
topology on F (Rn R)
the set of u.s.c. functions of Rn in R is a complete lattice for :
f g SG(f ) SG(g)
Algebraic dilation and erosion
complete lattice (T , )
Algebraic dilation:
(xi ) T , (i xi ) = i (xi )
Algebraic erosion:
(xi ) T , (i xi ) = i (xi )
Properties:
(0) = 0 (in P(E), 0 = )

(I) = I (in P(E), I = E)
increasing
increasing
in P(Rn ), (X) = xX ({x})
Adjunctions
(, ) adjunction on (T , ):
(x, y), (x) y x (y)
Properties:
(0) = 0 and (I) = I
increasing = algebraic erosion iff such that (, ) is an adjunction

V
= algebraic dilation and (x) = {y T , (y) x}
(, ) adjunction = algebraic erosion and = algebraic dilation

increasing = algebraic dilation iff such that (, ) is an adjunction
W
= algebraic erosion and (x) = {y T , (y) x}
Id
Id
=
=
= and =
Links with morphological operators
On the lattice of the subsets of Rn or Zn , with inclusion:

(X) = xX ({x})
+ invariance under translation B, (X) = D(X, B)

Same result on the lattice of functions.
Similar results for erosion.
Algebraic opening and closing
Algebraic opening: increasing, idempotent and anti-extensive
1 and 2 closings equivalence between:

1. 2 1
2. 1 2 = 2 1 = 1
3. Inv(1 ) Inv(2 )
Algebraic closing: increasing, idempotent and extensive

Examples: = and = with (, ) = adjunction
Invariance domain: Inv() = {x T , (x) = x}
W
opening (x) = {y Inv(), y x}
V
closing (x) = {y Inv(), x y}
W
(i ) openings i i opening
V
(i ) closings i i closing
1 and 2 openings equivalence between:
1. 1 2
2. 1 2 = 2 1 = 1
3. Inv(1 ) Inv(2 )
Algebraic filter theory
Filter = increasing and idempotent operator

Examples
W
openings and i i (anti-extensive filters)
V
closings and i i (extensive filters)
Theorem on filter composition and such that :

, , and are filters
Inv() = Inv() and Inv() = Inv()
is the smallest filter which is largest than
Example: alternate sequential filters
openings i and closings i such that:

i j j i Id i j
Theorem on filter composition mi = i i , ni = i i , ri = i i i and

si = i i i are filters
Alternate sequential filters:
Mi
mi mi1 ...m2 m1
Ni
ni ni1 ...n2 n1
Ri
ri ri1 ...r2 r1
Si
si si1 ...s2 s1
Property: i j Mj Mi = Mj , Nj Ni = Nj , ...
Morphological alternate sequential filters
(...(((fB1 )B1 )B2 )B2 )...Bn )Bn
(...(((fB1 )B1 )B2 )B2 )...Bn )Bn
(...(((fB1 )B1 )B2 )B2 )...Bn )Bn
Auto-dual filters
Operators which are independent of the local contrast, acting similarly on bright
and dark areas.
Example: morphological center

M edian[f (x), 1 (f )(x), 2 (f )(x)]
More generally, for operators {1 , 2 , ...n }: (Id i i ) i i

For instance 1 (f ) = (f ) = (f B )B , 2 = (f ) = (fB )B
Auto-dual filters
Operators which are independent of the local contrast, acting similarly on bright
and dark areas.
Example: morphological center

M edian[f (x), 1 (f )(x), 2 (f )(x)]
More generally, for operators {1 , 2 , ...n }: (Id i i ) i i

For instance 1 (f ) = (f ) = (f B )B , 2 = (f ) = (fB )B
O
F
centre
O
F
Morphological center: numerical example
Closing
Opening
Closing then opening
Opening then closing
morphological center
Closing then opening
Opening then closing
Hit-or-Miss Transformation
Structuring element: T = (T1 , T2 ), with T1 T2 =
HMT:
X T = E(X, T1 ) E(X C , T2 )
Hit-or-Miss Transformation
Structuring element: T = (T1 , T2 ), with T1 T2 =
HMT:
X T = E(X, T1 ) E(X C , T2 )
Thinning (if O T1 ):
X T =X \X T
Thickening (if O T2 ):
X T =X X T
For T 0 = (T2 , T1 ):
X T = (X C T 0 )C
HMT: examples
HMT: examples
Skeleton: requirements
compact representation of objects

thin lines
included and centered in the object
homotopic to the object
good representation of the geometry
invertible (reconstruction of the initial object)
Skeleton: continuous case

A: open set
s (A) = set of centers of maximal balls of A of radius
Skeleton:
r(A) =
s (A)
>0
Characterization:
s (A) =
[E(A, B ) \ [E(A, B )]B ]
>0
r(A) =
[ \
[E(A, B ) \ [E(A, B )]B ]
>0 >0
Reconstruction:
A=
D(s , B )
>0
Properties of the continuous skeleton
s (E0 (A)) = s+0 (A) r(E0 (A)) = >0 s (A)

no general formula for the skeleton of the dilation, opening or closing of a set
A 7 r(A) is l.s.c. from G in F
A connected r(A) connected
the skeleton is thin: its interior is empty
Skeleton: digital case
Direct transposition of the continuous case:

S(X) =
[E(X, Bn ) \ E(X, Bn )B ]
nN
Properties:
centers of digital maximal balls

reconstuction
but poor connectivity properties
Skeleton: digital case
Direct transposition of the continuous case:

S(X) =
[E(X, Bn ) \ E(X, Bn )B ]
nN
Properties:
centers of digital maximal balls

reconstuction
but poor connectivity properties
Skeleton from homotopic thinning

1
.
1
1
.
0
Properties:
perfect topology
no reconstruction
Centers of maximal ball vs thinning
Application in brain imaging

(PhD of Jean-Franois Mangin)






Geodesic operators
d (x,y)
X
Geodesic distance, conditional to X: dX
X
d(x,y)
if X is closed, there exits a geodesic arc

for any pair of points of X
unique if X is simply connected
X convex dX = d
Geodesic ball: BX (x, r) = {y X / dX (x, y) r}

Rq: BX (x, r) B(x, r)
Geodesic dilation:
DX (Y, Br ) = {x Rn / BX (x, r) Y 6= } = {x Rn / dX (x, Y ) r}
Geodesic erosion:
EX (Y, Br ) = {x Rn / BX (x, r) Y } = X \ DX (X \ Y, Br )
Geodesic opening and closing: by composition
Properties and reconstruction
Properties:
similar as in the Euclidean case

DX (Y, Br ) D(Y, Br )
DX (Y, Br ) =
n=1 [(Y
r
B)
n
X]n
Digital case:
DX (Y, Br ) = [D(Y, B1 ) X]r
Reconstruction:
[D(Y, B1 ) X] = DX
(Y )
= connected components of X which intersect Y
Binary reconstruction: example
Geodesic operators on functions

X1 X2 and Y1 Y2 DX1 (Y1 , Br ) DX2 (Y1 , Br ) DX2 (Y2 , Br )
Extension to functions, for f g, cut by cut:
[Dg (f, Br )] = Dg (f , Br )
(with f = {x, f (x) })
Digital case:
Dg (f, Br ) = [D(f, B1 ) g]r
Eg (f, Br ) = [E(f, B1 ) g]r
Numerical reconstruction of f (marker function) in g:
by dilation Dg (f, B ) = Dg (f ): opening
by erosion Eg (f, B ): closing

opening by reconstruction: Df (fB ) (flat areas whose contours are some
contours of the original image compression)
Numerical reconstruction: example
Opening by reconstruction: examples
Opening by reconstruction: examples
Union of openings by segments of length 20 and reconstruction
Application to alternate sequential filters
ASF with an hexagon (maximal size = 1)
ASF with segments (maximal size = 1)
Regional maxima
X regional maximum of f if
x X, f (x) = et X = CC(f )
Computation of regional maxima:
f Df (f 1)
h-maxima (grey level dynamics):
f Df (f h)
robust maxima
Regional maxima: example
Robust maxima: example
Skeleton by influence zones
X=
Xi
Influence zone of Xi in X C :
ZI(Xi ) = {x X C / d(x, Xi ) < d(x, X \ Xi )}
Skeleton by influence zones:
Skiz(X) = (
ZI(Xi ))C
= generalized Vorono diagram

Properties:
Skiz(X) Skel(X C )
Skiz is not necessarily connected (even if X C is)
Skeleton by influence zones: examples
Skeleton by influence zones: examples
Skeleton and skeleton by influence zones
Skiz(X) r(X C )
Skeleton and skeleton by influence zones
Skiz(X) r(X C )
Geodesic skeleton by influence zones
Y = i Yi
Geodesic influence zone of Yi conditionally to X :
ZIX (Yi ) = {x X, dX (x, Yi ) < dX (x, Y \ Yi )}
Geodesic skeleton by influence zones:
SKIZX (Y ) = X \
ZIX (Yi )
SKIZ(Y)
SKIZ X (Y)
Y1
Y1
Y2
Y2
X
Cortex segmentation
(PhD of Arnaud Cachia)
Cortex segmentation
(PhD of Arnaud Cachia)
Watersheds
catchment basins
watersheds
minimum
minimum
Watersheds: definition
Steepest descent:
Desc(x) = max{
f (x) f (y)
, y V (x)}
d(x, y)
Ramp of a path = (x0 , ...xn ):
Tf () =
n
X
d(xi1 , xi )Cost(xi1 , xi )
i=1
with
Desc(x)
Cost(x, y) =
Desc(y)
(Desc(x) + Desc(y))/2
if f (x) > f (y)

if f (y) > f (x)
if f (y) = f (x)
Watersheds: definition
Topographic distance
Tf (x, y) = inf{Tf (), = (x0 = x, x1 , ..., xn = y)}
(equals 0 on a plateau)
Catchment basin associated to the regional minimum Mi :
BV (Mi ) = {x, j 6= i, Tf (x, Mi ) < Tf (x, Mj )}
Watersheds:
LP E(f ) = [i BV (Mi )]C
Approach by immersion
Construction of the watersheds

f such that f (x) [hmin , hmax ], f h = {x, f (x) h}
Xhmin
f hmin
Xh+1
M inRegh+1 (f ) ZIf h+1 (Xh )
BV
Xhmax
LP E(f )
XhCmax
Illustration of the algorithm
Watersheds and oversegmentation
Watersheds and oversegmentation
Geodesic erosion
in order to impose markers
Watersheds constraint by markers

f : function on which watersheds should be applied
g: marker function (selects regional minima)
Reconstruction: Ef g (g, B ) (only the selected minima)
Separation of connected binary objects
And much more...

morphoEngAll2 PDF

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

morphoEngAll2 PDF

Загружено:

Авторское право:

Доступные форматы

Mathematical Morphology

Ecole Nationale Superieure

Mathematical Morphology p.1/84

J. Serra, Image Analysis and Mathematical Morphology, Academic Press,

P. Soille, Morphological Image Analysis, Springer-Verlag, Berlin, 1999.

Mathematical Morphology p.2/84

Shape or spatial relations?

Mathematical Morphology p.3/84

Simplifying and selecting relevant information...

Mathematical Morphology p.4/84

Origin: study of porous media

grey levels, colors

shape, geometry, topology

Mathematical Morphology p.5/84

Applications in numerous domains

Mathematical Morphology p.6/84

Mathematical Morphology p.7/84

Mathematical Morphology p.7/84

Four fundamental principles

1. Compatibility with translations

Mathematical Morphology p.8/84

Mathematical Morphology p.9/84

Mathematical Morphology p.10/84

extensive (X D(X, B)) if O B;

D(X Y, B) D(X, B) D(Y, B);

iterativity property: D[D(X, B), B 0 ] = D(X, B B 0 ).

Mathematical Morphology p.10/84

Mathematical Morphology p.11/84

Mathematical Morphology p.12/84

duality of erosion and dilation with respect to complementation:

E[(X Y ), B] E(X, B) E(Y, B);

iterativity property: E[E(X, B), B 0 ] = E(X, B B 0 ).

Mathematical Morphology p.12/84

Mathematical Morphology p.13/84

Links with distances

Mathematical Morphology p.14/84

Links with distances

Mathematical Morphology p.14/84

(or D[E(X, B), B])

Mathematical Morphology p.15/84

(or D[E(X, B), B])

Mathematical Morphology p.15/84

Mathematical Morphology p.16/84

(or E[D(X, B), B])

Mathematical Morphology p.17/84

(or E[D(X, B), B])

Mathematical Morphology p.17/84

Mathematical Morphology p.18/84

Mathematical Morphology p.19/84

From sets to functions

subgraph of a function on Rn = subset of Rn+1

functional equivalents of set operations:

Mathematical Morphology p.20/84

Mathematical Morphology p.21/84

Properties of functional dilation:

Mathematical Morphology p.21/84

Example of functional dilation

Mathematical Morphology p.22/84

x Rn , E(f, B)(x) = inf{f (y) / y Bx }

Mathematical Morphology p.23/84

x Rn , E(f, B)(x) = inf{f (y) / y Bx }

Properties of functional erosion:

functional dilation and erosion are dual operators;