International Journal of Information and Computer Science

IJICS Volume 1, Issue 7, October 2012 PP. 178-187

ISSN(online) 2161-5381
ISSN(print) 2161-6450
www.iji-cs.org

Cubical Homology in Digital Images

Ismet Karaca1 , Ozgur Ege2
Department of Mathematics, Ege University, Bornova, Izmir 35100, TURKEY
1
ismet.karaca@ege.edu.tr, 2ege-ozgur87@hotmail.com

1, 2

(Abstract) In this article we study the digital cubical homology groups of digital images which are based on the cubical
homology groups of topological spaces in algebraic topology. We investigate some fundamental properties of cubical homology
groups of digital images. We also calculate cubical homology groups of certain 2-dimensional and 3-dimensional digital images.
We give a relation between digital simplicial homology groups and digital cubical homology groups. Moreover we show that the
Mayer-Vietoris Theorem need not be hold in digital images.
Keywords: Digital Topology; Digital Cubical Set; Digital Cubical Homology Group; Euler Characteristic; Mayer-Vietoris
Theorem

1. INTRODUCTION
Topological invariants are useful in many applications related
to digital imaging and geometric modeling, and homology is a
classical one. The higher homotopy groups and homology
groups are useful algebraic tools in a large number of
topological problems, and are computational tools of
algebraic topology. The digital fundamental group is a nice
tool to classify the digital images with k-adjacency relations
but it does not yield completely information in a great class of
explicit problems. We need to establish a new algebraic
scructure which is called the digital homology groups in order
to classify the various digital images with k-adjacency
relations. The digital cubical homology can be an important
tool to classify digital images. Many researchers (Kaczynski,
Mischaikow, Mrozek, Allili, Tannenbaum, Kalies, Watson,
Pilarczyk, Zelazna and Kot) have studied cubical homology.
Kalies, Mischaikow and Watson [8] introduce a method for
computing the homology groups of cellular complexes
composed of cubes. The algorithm used in the homology
computations is based on a local reduction procedure, and
they give an estimate of its computational complexity. This
estimate is rigorous in two dimensions, and they conjecture its
validity in higher dimensions.
Allili, Mischaikow and Tannenbaum [1] combine a new
method combinatorial topology to compute the number of
connected components and holes of objects in a given image,
and fast segmentation methods to extract the objects. Their
computational method for determining the homology groups
is based on a reduction process of the size of the chain
complex by local simplification in such a way that the
homology is preserved at each step.
Kaczynski, Mischaikow and Mrozek [7] present cubical
sets and the algebra of cubical sets. Also, they define cubical
homology and investigate its most elementary properties.
They present a computational approach to homology with the

hope that such a theory will prove beneficial to the analysis

and understanding of todays complex geometric challenges.
They give a relation between cubical complexes and image
data. They compare with cubical and simplicial complexes.
Cubical complexes [7] have several nice properties that
simplicial complexes do not share. Images and numerical
computations naturally lead to cubical sets. Subdividing these
cubes to obtain a triangulation is at this point artificial and
increases the size of data significiantly. For example, it
requires n! simplices to triangulate a single n-dimensional
cube. Because cubical complexes are so rigid, they can be
stored with a minimal amount of information. For example,
an elementary cube can be described using one vertex.
Specifying a simplex typically requires knowledge of all the
vertices. A product of elementary cubes is an elementary cube,
but a product of simplices is not a simplex. Cubes are more
convenient than simplices for constructing products. It's easier
to construct a cubic chain on a product X Y given cubic
chains on X and Y. Thus cubes are useful for local-to-global
problems. Also cubes have a nice tensor product property and
this is crucial for obtaining some homotopy classification
results.
The main areas in which computational cubical homology
are applications in digital image processing/analysis,
dynamical systems and medicine/scructural biology. The
main applications in medicine/scructural biology also concern
digital imagery. For example, cubical homology has been
used to extract topological information, e.g. cavities, directly
from raw MRI/MRA-data. This usefulness of this is that
creating an image is costly in terms of time and memory used
by the computer and thus if we are only interested in
topological information we only need to use cubical
homology which is much cheaper and thus allows for more
detailed calculations under the same conditions.
Kot [9] presents an algorithm based on the simplification of
n
the boundaries and coboundaries of cubes in R . This

IJICS Volume 1, Issue 7 October 2012 PP. 178-187 www.iji-cs.org Science and Engineering Publishing Company
- 160 -

International Journal of Information and Computer Science

algorithm is an extension of the results presented in [8]. The

algorithm obtained allows to compute the homology groups
for two-dimensional cubical sets.
Mrozek, Pilarczyk and Zelazna [10] present a new
reduction algorithm for the efficient computation of the
homology of a cubical set. The algorithm is based on
constructing a possible large acyclic subspace, and then
computing the relative homology.
This paper is concerned with setting up more algebraic
invariants for a digital image with k-adjacency. The purpose
of this paper is to give complete algebraic presentation of
cubical homology group of any objects in digital image. Some
results related to cubical homology groups of a digital image
are given. We also calculate the cubical homology group of
certain 2-dimensional and 3-dimensional digital images. The
paper is organized as follows: In Section 2 we introduce the
general notions of digital images with k-adjacency relations.
In Section 3 we define a cubical digital image, we introduce
the algebra of digital cubical sets and we define the digital
boundary operator. In Section 4 we define the cubical
homology groups of digital images, we give some examples
about cubical homology groups of digital images, we compute
the digital cubical homology groups of a minimal simple
curve that its digital simplicial homology groups are found by
Boxer, Karaca and Oztel [4], we conclude that digital
simplicial homology groups and digital cubical homology
groups of a digital image need not be isomorphic, we define
Euler characteristic of a digital cubical set and compute Euler
characteristic of MSS 6' . In Section 5 we show that the
Mayer-Vietoris Theorem need not be hold in digital images.

2. PRELIMINARIES
n

Let Z be the set of integers. Then Z is the set of lattice

points in the n-dimensional Euclidean space. A (binary)
n
digital image is a finite subset of Z with an adjacency
relation. A variety of adjacency relations are used in the study
of digital images, including the following.
Definition

2.1

[3]

For

positive

integer

with

1 l n and distinct two points

p = ( p1 , p 2 ,..., p n ), q = (q1 , q 2 ,..., q n ) Z n

IJICS
2

(2) Two points p and q in Z are 4-adjacent if they are

8-adjacent and differ in exactly one coordinate.
3

(3) Two points p and q in Z are 26-adjacent if they are

distinct and differ by at most 1 in each coordinate.
3

(4) Two points p and q in Z are 18-adjacent if they are

26-adjacent and differ in at most two coordinate.
3

(5) Two points p and q in Z are 6-adjacent if they are

18-adjacent and differ in exactly one coordinate.
(6) Let k {4,8,6,18,26} . A k - neighbor of a lattice
point p is k -adjacent to p. More general adjacency
relations are studied in [6].
Definition 2.2 [2] Let a, b Z with a < b . A set of the
form

[a, b] Z = {z Z a z b}

is called a digital interval.

Definition 2.2 [11] Let k be an adjacency relation defined
n
n
on Z . A digital image X Z is k -connected if and
only if for every pair of different points x, y X , there is a
set {x 0 , x1 ,..., x r } of points of a digital image X such that
x = x0 , y = x r and xi and xi +1 are k -neighbors where
i = 0,1,..., r 1 . A k -component of a digital image X is a
maximal k -connected subset of X.
n
n
Let X Z 0 and Y Z 1 be digital images with
k 0 -adjacency and k1 -adjacency respectively. A function
f : X Y is said to be ( k 0 , k1 )-continuous [3] if for
every k 0 -connected subset U of X, f(U) is a k1 -connected
subset of Y . Such a function is called digitally continuous.
Preposition 2.4 [3] Let X Z 0 and Y Z 1 be digital
images with k 0 -adjacency and k1 -adjacency respectively.
Then the function f : X Y is ( k 0 , k1 )-continuous if and
only if for every k 0 -adjacent points {x 0 , x1 } of X, either
f ( x0 ) = f ( x1 ) or f ( x0 ) and f ( x1 ) are a k1 -adjacent
in Y.
By a digital k -path from x to y in a digital image X, we
mean a (2, k )-continuous function f : [0, m] Z X such
that f(0) = x and f(m) = y. If f(0) = f(m) then the k -path is
said to be closed. A simple closed k -path is considered as a
sequence { f (0), f (1),..., f ( m 1)} of images of the
k -path f : [0, m] Z X such that f(i) and f(j) are
k -adjacent if and only if j = i mod m.
n

(2.1)

p and q are k l - adjacent if

(1) there are at most l indices i such that
pi qi = 1 and
(2)
for
all
other
indices
j
such
that
p j q j 1, p j = q j .
Note that the notation k l is the number of points
q Z n that are adjacent to a given point p Z n in this
sense. From Definition 2.1, we have the following;
2
(1) Two points p and q in Z are 8-adjacent if they are
distinct and differ by at most 1 in each coordinate.

(2.2)

IJICS Volume 1, Issue 7 October 2012 PP. 178-187 www.iji-cs.org Science and Engineering Publishing Company
- 161 -

International Journal of Information and Computer Science

Let X Z 0 and Y Z 1 be digital images with

k 0 -adjacency and k1 -adjacency respectively. A function
f : X Y is a ( k 0 , k1 )-homeomorphism (see [2]) if f is
( k 0 , k1 )-continuous and bijective and further
f 1 : X Y is ( k1 , k 0 )-continuous. We use the notation
X ( k0 ,k1 ) Y .
n

IJICS

Example 3.2 Consider the digital image

X = [0,1]Z [2,3]Z [4,5]Z Z 3 with 26-adjacency.

This is an elementary digital cube and hence, is a digital
cubical set.

3. DIGITAL CUBICAL SETS

K 2 ( X ,26) = {[0] [2,3] [4,5], [1] [2,3] [4,5],

We benefit from [7] to study cubical homology in digital

images.

[0,1] [2] [4,5], [0,1] [3] [4,5], [0,1] [2,3] [4], [0,1] [2,3] [5]},

Definition 3.1 [7] An elementary digital interval is a digital

interval I Z of the form

I = [a, a + 1]Z or I = [a, a ]Z

(3.1)

for some a Z. If a digital interval contains only one point,

[ ]Z = [a, a ]Z

then we write a

to simplify the notation.

Elementary digital intervals that consist of a single point are

degenerate, whereas those of length 1 are nondegenerate. An
elementary digital cube Q with 2n-adjacency is a finite
product of elementary digital intervals, that is,

Q = I 1 I 2 I n Z n

(3.2)

Where each I i is an elementary digital interval. The

number of nondegenerate components in Q is called the

K 1 ( X ,26) = {[0] [2] [4,5], [0] [3] [4,5]

[0] [2,3] [4], [0] [2,3] [5], [1] [2] [4,5],

[1] [3] [4,5], [1] [2,3] [4], [1] [2,3] [5],
[0,1] [2] [4], [0,1] [2] [5], [0,1] [3] [4],
[0,1] [3] [5]}
K 0 ( X ,26 ) = {[0] [2] [4], [0] [2] [5], [0] [3] [4],

[1] [2] [5], [0] [3] [5], [1] [2] [4], [1] [3] [4],

[1] [3] [5]}.

dimension of Q and it is denoted by dim Q. The set of all

n

( ) and

elementary digital cubes in Z is denoted by K Z

( ).

those of dimension q by K q Z

We now present an algebraic structure for digital cubical

sets arbitrary dimension. With each elementary digital q-cube

Q K q with -adjacency we associate Q called an

-adjacency. The set

Z with -adjacency

A digital image (X, 2n) Z is cubical if (X, 2n) can be

written as a finite union of elementary digital cubes. If (X,
n
2n) Z is a digital cubical set, then we use the following
notation:
(3.3)
K ( X ,2n ) = {Q K : Q ( X ,2n )}

elementary digital q-chain of Z with

and

and the set of all elementary digital chains of Z is given by

of all elementary digital q-chains of

is denoted by

K q = Q : Q K q

(3.5)
n

K q ( X ,2n ) = {Q K ( X ,2n ) : dim Q = q} (3.4)

K = K q

The elements of K 0 ( X , k ) are the vertices of ( X , k ) and the

elements of K 1 ( X , k ) are the edges of

( X , k ) . More

generally, the elements of K q ( X , k ) are the digital q-cubes

of (X,k).

(3.6)

q =0

, Q ,...Q K of
Given any finite collection Q
1
2
n
q
q-dimensional elementary digital chains, we can consider
sums of the form

c = 1Q 1 + 2 Q 2 + ... + n Q n

IJICS Volume 1, Issue 7 October 2012 PP. 178-187 www.iji-cs.org Science and Engineering Publishing Company
- 162 -

(3.7)

International Journal of Information and Computer Science

where

are arbitrary integers. If all the

= 0. The set of digital q-chains with

IJICS

0, Q = [a ]
q Q =
[a + 1] [a ], Q = [a, a + 1]

()

i = 0, , then c

-adjacency is denoted

by C q . C q is an abelian group and it is a free abelian

Q K q , define

group with basis K q . For each

Let I = I 1 (Q ) and P = I 2 (Q ) ... I n (Q ) .Then

Q = IP .Define

()

dim I
q Q = q1 IP + ( 1) I q2 P

Q : K qn by
1, P = Q
Q (P ) =
0, P Q

(3.8)

where q1 = dim I and q 2 = dim P. Finally, we extend the

c = 1Q 1 + 2 Q 2 + ... + m Q m , then
q (c ) = 1 q Q 1 + 2 q Q 2 + ... + m q Q m

-adjacency is the free abelian group

chains of Z with

generated by the elementary digital chains of K q . In

K q and K q .

dim c1

q : C q ( X ) C q1 ( X )
(3.9)

Let (X, )

q C q ( X ) C q1 ( X ) .

c 2 C q22 by

4.

c1 , P c 2 , Q P Q

(3.10)

q2

The chain c1 c 2 is called the digital cubical product of c1

and c 2 .
Definition 3.5 [7] Given q , the cubical boundary
operator

q : C q C q 1

(3.11)

is a homomorphism of free abelian groups, which is defined

K
for an elementary chain Q
q

(3.16)

be a digital cubical set. Then

This definition extends to arbitrary chains c1 C q11 and

,Q

(3.15)

Definition 3.8 [7] The digital boundary operator for the

digital cubical set (X, ) is defined to be

P Q = P Q

q1

c1 q c 2

Proposition 3.7 [7] q q +1 = 0 for all q 0.

P K q1 and Q K q2 , set

q c1 c 2 = q c1 c 2 + ( 1)

Definition 3.4 [7] Given two elementary digital cubes

(3.14)

Proposition 3.6 [7] Let c1 and c 2 be digital cubical

chains with -adjacency; then

particular, K q is the basis for C q .

Note that there is a bijection between

(3.13)

definition to all chains by linearity; that is, if

Definition 3.3 [7] The group C q of q-dimensional digital

c1 c 2 =

(3.12)

CUBICAL HOMOLOGY IN
IMAGES

DIGITAL

Let X be a digital cubical set [7] with

The kernel of

-adjacency.

q : C q ( X ) C q1 ( X )

(4.1)

is called the group of digitally cubical q-cycles in (X, ) and

denoted by Z q ( X ) . The image of

q +1 : C q+1 ( X ) C q ( X )

(4.2)

is called the group of digitally cubical q-boundaries in (X, )

and denoted by Bq ( X ) .
Since Proposition 3.7, each digitally cubical q-boundary of
digitally cubical (q+1)-chains is again a digitally cubical

q-cycle, that is, Bq ( X ) is a normal subgroup of Z q ( X )

IJICS Volume 1, Issue 7 October 2012 PP. 178-187 www.iji-cs.org Science and Engineering Publishing Company
- 163 -

International Journal of Information and Computer Science

for each q 0 . To give the nontrivial cycles, we introduce

equivalence relation. We say that two cycles
z1 , z 2 Z q ( X ) are homologous and we write z1 z 2 if
z1 z 2 Z q ( X ) is a boundary in X, that is,
z1 z 2 Bq ( X ) . The equivalence classes are elements of

the quotient group Z q ( X ) Bq ( X ) .

Definition 4.1 [7] The qth digital cubical homology group is
the quotient group

IJICS

be a digital cubical set with 4-adjacency. The sets of

elementary digital cubes are

0 ( X ,4 ) = {[0] [0], [0] [1], [1] [0], [1] [1]}

1 ( X ,4 ) = {[0] [0,1], [1] [0,1], [0,1] [0], [0,1] [1]}
Thus the bases for the sets of chains are

H q ( X ) = Z q ( X ) Bq ( X ).
The homology of X is the collection of all homology groups
of X, that is,

H ( X ) = H q ( X ) q
Given z Z q ( X ) ,

[z ] H q ( X )

(4.3)

Example 4.2 [7] Let X = . Then C q ( X ) = 0 for all q and

hence
q = 0,1,2,...
(4.4)
H q ( X ) = 0

Example 4.3 [7] Let X = {x 0 }

be a digital cubical set

= {[0][0], [0][1], [1][0], [1][1]}

K 1 ( X ,4) = {[0][0,1], [1][0,1], [0,1][0], [0,1][1]}

= {[0][0 , 1], [1][0 , 1], [0 , 1][0], [0 , 1][1]}.

is the homology class

of z in X.

K 0 ( X ,4) = {[0][0], [0][1], [1][0], [1][1]}

To simplification, in 0 ( X ,4 )

a1 = [0][0], a 2 = [0][1], a3 = [1][0], a 4 = [1][1],

in 1 ( X ,4 )

consisting of a single point with adjacency. Then

b1 = [0][0 , 1], b2 = [1][0 , 1], b3 = [0 , 1][0], b4 = [0 , 1][1].

x0 = [a1 ] [a 2 ] ... [a n ]. Thus

Then chain groups are

, k = 0
Cq (X ) =
0, k 0

(4.5)

0
2
1
C14 ( X )
C 04 ( X )
0
0

C1 ( X ) = 0, B0 ( X ) = 0 , and therefore, H 0 ( X ) .
Since C q ( X ) = 0 for all q 1 , H q ( X ) = 0 for all

Thus, we get the following short sequence :

Furthermore Z 0 ( X ) C 0 ( X ) = . Since

C 04 ( X ) = {a1 , a 2 , a 3 , a 4 } and C14 ( X ) = b1 , b2 , b3 , b4 .

To compute the boundary operator we need to compute the

boundary of the basis elements.

()
(b ) = a

1 b1 = a1 + a 2
1

q 1 . Therefore,

(4.7)

+ a 3

( )
(b ) = a

1 b2 = a 3 + a 4
1

+ a 4

(4.8)

To determine 1 ( X ) , we need to know Ker 1 , that is, we

need to solve the equation
4

, k = 0
H q ( X ) =
0, k 0.

(4.6)

Example 4.4 [7] Let

X = [0] [0,1] [1] [0,1] [0,1] [0]

[0,1] [1]

()
( )
+ (b )+ (b )
)(b ) + ( + )(b )

1 1b1 + 2 b2 + 3 b3 + 4 b4 = 1 1 b1 + 2 1 b2
3 1

()
( )

= ( 1 3 ) b1 + (1
+ ( + ) b .
2

4 1

IJICS Volume 1, Issue 7 October 2012 PP. 178-187 www.iji-cs.org Science and Engineering Publishing Company
- 164 -

International Journal of Information and Computer Science

Solving the equation

( 1 3 )(b1 ) + (1 4 )(b2 ) + ( 2 + 3 )(b3 ) + ( 2 + 4 )(b4 ) = 0

we have

1 = 2 = 3 = 4 .

Hence

{(

Since C 2 ( X ) = 0, B1 ( X ) = 0 and hence

4

4
1

c 4 = (0,0,1), c5 = (1,0,1), c6 = (1,1,1), c7 = (0,1,1)},

(X ) = Z (X ) .
4
1

We turn to computing H
solution to the equation

4
0

( X ) . First observe that there is no

()

a1 B04 ( X ) . On the other hand,

1 b1 = a1 + a 2

1 b1 + b4
Thus,

( )
b ) = a + a .
1

MSS 6' = [0,1] [0,1] [0,1] 3

, q = 0,2
H q6 MSS 6' =
0, q 0,2.

1 b1 + b4 = a1 + a 4
2

That is,

is a digital cubical set with 6-adjacency (see figure 1), then its
digital cubical homology groups are

1 1b1 + 2 b2 + 3 b3 + 4 b4 = b1 .
This implies that

, k = 0,1
H k4 ( X ) =
0, k 0,1.
MSS 6' = {c 0 = (0,0,0), c1 = (1,0,0), c 2 = (1,1,0), c 3 = (0,1,0 ),

14 ( X ) = b1 b2 b3 b4 .
4

that

Theorem 4.5 If

IJICS

{a1 a 2 , a1 a 4 , a1 a 3 } B04 ( X ) .

In particular, all the elementary chains are homologous, that

is,

a1 a 2 a 3 a 4 .
Now consider an arbitrary chain

z C 04 ( X ) .Then

z = 1 a1 + 2 a 2 + 3 a 3 + 4 a 4 .

Figure 1:5 MSS 6

'

So on the level of homology

[z ]X = [ 1a1 + 2 a 2 + 3 a 3 + 4 a 4 ]X

'

Proof. The sets of elementary digital cubes of MSS 6 are

2 (MSS 6' ,6) = {[0,1] [0,1] [0], [0,1] [0] [0,1], [0,1] [1] [0,1]

= 1 [a1 ]X + 2 [a 2 ]X + 3 [a3 ]X + 4 [a 4 ]X
= ( 1 + 2 + 3 + 4 )[a1 ]X

(4.9)

where the last equality comes from that fact that all the
elementary chains are homologous. Therefore, we can think
of every element of H q ( X ) = Z q ( X ) Bq ( X ) as being

generated by a1 and thus H 0 ( X ) .We have proven

[1] [0,1] [0,1], [0] [0,1] [0,1], [0,1] [0,1] [1]}

1 (MSS 6' ,6) = {[0,1] [0] [0], [0] [0] [0,1], [0] [0,1] [0],

[1] [0,1] [0], [1] [0] [0,1], [1] [1] [0,1], [0,1] [1] [0],
[0] [1] [0,1], [0,1] [0] [1], [0] [0,1] [1], [1] [0,1] [1],

IJICS Volume 1, Issue 7 October 2012 PP. 178-187 www.iji-cs.org Science and Engineering Publishing Company
- 165 -

International Journal of Information and Computer Science

IJICS

[0,1] [1] [1]}

c 0 = [0][0][0], c1 = [1][0][0],

0 (MSS ,6) = {[0] [0] [0], [1] [0] [0], [1] [1] [0],

c 2 = [1][1][0], c3 = [0][1][0],

[0] [1] [0], [0] [0] [1], [1] [0] [1], [1] [1] [1],

c 4 = [0][0][1], c5 = [1][0][1],

'
6

c 6 = [1][1][1], c 7 = [0][1][1].

[0] [1] [1]}.

Then chain groups are

) {

C 26 MSS 6' = {e0 ,e1 , e2 , e3 , e 4 , e5 }

Thus the bases for the sets of chains are

C16 MSS 6' = d 0 ,d1 , d 2 , d 3 , d 4 , d 5 , d 6 , d 7 , d 8 , d 9 , d10 , d11

C 06 MSS 6' = {c 0 ,c1 , c 2 , c3 , c 4 , c5 , c 6 , c 7 }.

Thus, we get the following short sequence :

3
2
1
0
C 26 MSS 6'
C16 MSS 6'

0
C06 MSS 6'
0

(4.10)

To compute the boundary operator we need to compute the

boundary of the basis elements.

To simplify the notation, in 2 MSS 6' ,6

2 ( e0 ) = d3 + d6 d4 d0

2 (e1 ) = d 0 + d1 d8 d 4

2 (e2 ) = d 6 + d 7 d11 d5

e0 = [0 , 1][0 , 1][0], e1 = [0 , 1][0][0 , 1],

e2 = [0 , 1][1][0 , 1], e3 = [1][0 , 1][0 , 1],

e4 = [0][0 , 1][0 , 1], e5 = [0 , 1][0 , 1][1],

in 1 MSS 6' ,6

2 (e4 ) = d 2 + d 7 d9 d1

d 0 = [0 , 1][0][0], d1 = [0][0][0 , 1],

d 2 = [0][0 , 1][0], d 3 = [1][0 , 1][0], ,

d 4 = [1][0][0 , 1], d 5 = [1][1][0 , 1],

d 6 = [0 , 1][1][0], d 7 = [0][1][0 , 1],

d 8 = [0 , 1][0][1], d 9 = [0][0 , 1][1],

d10 = [1][0 , 1][1], d11 = [0 , 1][1][1],

in 0 MSS 6' ,6

2 (e3 ) = d3 + d5 d10 d 4

2 (e5 ) = d10 + d11 d9 d8 .

To determine 2 MSS 6 , we need to know Ker 2 , that

is, we need to solve the equation
6

'

2 ( a 0 e0 + a1 e1 + a 2 e 2 + a 3 e3 + + a 4 e 4 + a 5 e5 ) =
( a 0 + a1 )d 0 + (a1 a 4 )d1 + ( a 0 + a 4 )d 2 + (a 0 + a 3 )d 3
+ ( a a )d + (a + a )d + (a + a )d
1

+ (a 2 + a 4 )d 7 + (a1 a 5 )d8
+ (a 4 a 5 )d 9 + (a 3 + a 5 )d10 + (a 2 + a 5 )d11 .

IJICS Volume 1, Issue 7 October 2012 PP. 178-187 www.iji-cs.org Science and Engineering Publishing Company
- 166 -

International Journal of Information and Computer Science

Solving the equation

IJICS

a 4 = a 0 a3

a 2 = a 0 a1

(a 0 + a1 )d 0 + (a1 a 4 )d1 + (a 0 + a 4 )d 2 + (a 0 + a 3 )d 3
+ (a a )d + ( a + a )d + (a + a )d

a 7 = a 0 a1 + a3 a5

+ (a 2 + a 4 )d 7 + ( a1 a 5 )d8
+ ( a 4 a 5 )d 9 + ( a 3 + a 5 )d10 + ( a 2 + a 5 )d11 = 0

a10 = a 0 a1 + a3 a5 + a8

Hence we get

Z 26 ( MSS 6' ) = {a (e0 + e1 e 2 e3 + e 4 e5 : a i Z } Z .

Since B2 ( MSS 6 ) = {0}, it follows that H 2 ( MSS 6 ) Z .
6

a11 = a5 a10 = a 0 + a3 a5 a8 .
Hence we get

a 0 = a1 = a 2 = a3 = a 4 = a5 .

'

a9 = a1 a8

we have

a6 = a3 + a5

'

Z 16 ( MSS 6' ) = {a 0 d 0 + a1 d1 + (a 0 a1 )d 2 + a3 d3 +
(a 0 a3 )d 4 + a5 d5 + ( a3 + a5 )d 6 + (a 0 a1 + a3 a5 )d 7
+ a d + (a a )d + (a a + a a + a )d +
8

To compute the boundary operator we need to compute the

10

(a 0 + a3 a5 a8 )d11 : ai Z } Z 5 .
'

We now compute B1 ( MSS 6 ). From the equation (4.11),

boundary of the basis elements.

we have

1 (d0 ) = c4 c0

1 (d 7 ) = c7 c3

1 (d 2 ) = c3 c0

1 (d3 ) = c7 c4

1 (d 4 ) = c5 c1

1 (d5 ) = c6 c2

B16 ( MSS 6' ) = {(k 0 + k1 )d0 + k 0 d1 + k1d2 +

( k1 k 3 k 4 )d3 + ( k 0 + k 3 + k 4 )d 4 + k 2 d5 +
( k k k k )d + ( k + k + k )d +

1 (d 6 ) = c2 c1

1 (d 7 ) = c6 c5

1 (d8 ) = c1 c0

1 (d9 ) = c5 c4

1 (d10 ) = c2 c3

'

'

Since B1 ( MSS 6 ) = Z 1 ( MSS 6 ), it follows that

H 16 ( MSS 6' ) {0}.

1 (a 0 d 0 + a1 d1 + a 2 d 2 + a 3 d 3 + + a 4 d 4 + a 5 d 5 +
a d + a d + a d + a d + a d + a d ) =

We now compute H 0 ( MSS 6 ).

'

10

10

: k i Z , i = 0,1,2,3,4} Z 5 .

To determine Z 1 ( MSS 6 ), we need to know Ker 1 ,

that is, we need to solve the equation

( k 0 + k 3 )d8 + k 3d9 + k 4 d10 + ( k 2 + k 4 )d11

1 (d11 ) = c6 c7

11 11

( a 0 a1 a 2 )c 0 + (a 0 a 3 a 4 )c1 + (a 3 a 5 + a 6 )c 2
+ (a 2 a 6 a 7 )c3 + (a1 a8 a10 )c 4 + (a 4 a 9 + a11 )c5
+ (a 5 + a 9 + a10 )c6 + (a 7 + a8 a10 )c7
(4.11)
Solving the equation

( a0 a1 a 2 )c0 + ( a0 a3 a 4 )c1 + ( a3 a5 + a6 )c2

+ ( a 2 a6 a7 )c3 + ( a1 a8 a10 )c4

'

Z 06 ( MSS 6' ) = {a 0 c0 + a1c1 + a 2 c2 + a3 c3 + a 4 c4 + a5 c5 +

a 6 c6 + a 7 c7 : ai Z } Z 8 .
Any 0-cycle

u 0 = a 0 c0 + a1c1 + a 2 c2 + a3 c3 + a 4 c4 + a5 c5 +
a 6 c6 + a 7 c7
can be written as

+ ( a 4 a9 + a11 )c5 + ( a5 + a9 + a10 )c6

+ ( a7 + a8 a10 )c7 = 0
we have
IJICS Volume 1, Issue 7 October 2012 PP. 178-187 www.iji-cs.org Science and Engineering Publishing Company
- 167 -

International Journal of Information and Computer Science

u 0 = 1 ((a1 + a 2 + a5 + a6 )d0 + a 2 d1 + (a3 + a7 )d3 + a 4 d 4

7

+ (a5 + a6 )d5 + a6 d6 + a7 d7 ) + ai c0 .

IJICS
'

Euler characteristic of MSS 6 was found that

( MSS 6' ,6) = 4(see [4]).

Then from Example 4.8, we

conclude that Euler characteristic of a digital image which

i =0

depends on whether cubical or not simplicial homology

Thus, u 0 is homologous to the 0-chain

groups may be different.

(a 0 + a1 + a 2 + a3 + a 4 + a5 + a 6 + a 7 )c0 .
Therefore, H 0 ( MSS 6 ) Z . Thus we have the required
6

'

Corollary 4.9 Euler characteristic of a digital image due to

digital cubical homology groups and its Euler characteristic
due to digital simplicial homology groups need not be the
same.

5. THE MAYER-VIETORIS THEOREM NEED

NOT BE HOLD IN DIGITAL IMAGES

result:

Z , q = 0,2
H q6 ( MSS 6' ) =
0, q 0,2.

Theorem 5.1 (Mayer-Vietoris) [7]. Let X be a cubical set.

Let X 1 and X 2 be cubical subsets of

'

X such that

The digital simplicial homology groups of MSS 6 are found

X = X 1 X 2 . Then there is a long exact sequence for all

that

n values
Z , q = 0

H q6 ( MSS 6' ) = Z 5 , q = 1
0, q 0,1

H n (X1 X 2 ) H n (X1) H n (X 2 )

(5.1)

H n ( X ) H n 1 ( X 1 X 2 )

Proposition 5.2 The Mayer-Vietoris theorem needn't be

hold in digital images.

(see [4]). Then we can state the following.

Proposition 4.6 The cubical and simplicial homology groups
of a digital image need not be isomorphic.

Proof Let X = [0,1] Z [0,1] Z Z

be a digital image

with 4-adjacency. Then

We now define Euler characteristic for digital cubical sets.

Z , n = 0,1
H n4 ( X ) =
0, n 0,1.

Definition 4.7 Let ( X , k ) be a digital cubical set.

p = rankH pk ( X )

(4.12)

(5.2)

Let

X 1 = [0,1] Z [0] Z [1] Z [0,1] Z X and

is called the pth Betti number of ( X , k ). The Euler

X 2 = [0,1] Z [0] Z [0] Z [0,1] Z X and then

characteristic of a digital cubical set ( X , k ) is the

X = X 1 X 2 . Also, X 1 X 2 = [0,1] Z [0] Z . We

alternating sum of its Betti numbers (see [7]),

know that

( X , k ) = (1) p .
p

p =0

Example 4.8 We can compute Euler characteristic of

MSS 6' in Teorem 4.5 by using Definition 4.7. From
Teorem 4.5,

( MSS ,6) = rankH ( MSS ) rankH ( MSS ) +

'
6

6
0

'
6

rankH 26 ( MSS 6' ) = 1 0 + 1 = 2.

6
1

'
6

Z , n = 0
H n4 ( X 1 X 2 ) = H n4 ( X 1 ) = H n4 ( X 2 ) =
0, n 0
(5.3)
Moreover, H 0 ( X ) = Z since
4

X is 4-connected digital

image. So we get following exact sequence :

IJICS Volume 1, Issue 7 October 2012 PP. 178-187 www.iji-cs.org Science and Engineering Publishing Company
- 168 -

International Journal of Information and Computer Science

(5.4)

From exactness of this sequence, we get H q ( X ) = 0 for

4

q 2.
k:Z Z Z
c k (c ) = (c, c )

(5.5)

Kerk = 0 and Im j = 0 because Im j = Kerk .

From first isomorphism theorem,

and we find that Kerj = H 1 ( X ) since

4

Sang-Eon Han, Digital fundamental group and Euler

characteristic of a connected sum of digital closed surfaces,
Information Sciences 177 (2007) no. 16, 3314-3326.
[6] G.T. Herman, Oriented surfaces in digital spaces, CVGIP:
Graphical Models and Image Processing 55 (1993), 381-396.
[7] T. Kaczynski, K. Mischaikow, et M. Mrozek, Computational
Homology, Appl. Math. Sci. Vol. 157, Springer Verlag , NY
2004.
[8] W. Kalies, K. Mischaikow, and G. Watson, Cubical
Approximation and Computation of Homology, in: Conley
Index Theory, Banach Center Publications 47(1999), 115131.
n
[9] P. Kot, Homology Calculation of Cu bical Complexes in R ,
Computational Methods In Science and Technology 12(2),
115-121 (2006).
[10] M. Mrozek, P. Pilarczyk, N. Zelazna, Homology algorithm
based on acyclic subspace, Comput. Math. Appl., Vol. 55, No.
11 (2008), 2395-2412.
[11] A. Rosenfeld, Continuous functions on digital pictures, Pattern
Recognition Letters 4 (1986), 177-184.
[5]

j
k
0 H 14 ( X )

Z Z Z 0

H 14 ( X ) / Kerj Im j

IJICS

(5.6)

Im j = 0. As

Im i = 0 and Im i = Kerj , we get Kerj = 0. So

Author Introduction
Ismet Karaca- was born in Afyon,
Turkey on January 5th, 1969. He received a

H ( X ) 0 but H ( X ) Z . We find a contradiction.

4
1

4
1

Bachelor's degree in Mathematics from

Anadolu University in Turkey, a Master's

So the Mayer-Vietoris theorem needn't be hold in digital

images.

6. CONCLUSION

in Mathematics from the university of

Miami, and a PhD in Mathematics from
Lehigh University. He is a Professor of
Mathematics at Ege University in Izmir,

The goal of this article is to compute cubical homology groups of

certain digital images. Although the cubical homology groups and
the simplicial homology groups of a topological space are
isomorphic in algebraic topology, we conclude that the cubical and
simplicial homology groups of a digital image need not be
isomorphic. Then we define Euler characteristic of a digital cubical
set and compute Euler characteristic of a digital cubical set. At the
same time we show that the Mayer-Vietoris Theorem need not be
hold in digital images.

TURKEY. Dr. Karaca's research interests include homotopy theory,

steenrod algebra, and digital topology. Prof. Karaca is a member of
American Mathematical Society.
Ozgur Ege- was born in Izmir, Turkey
on June 15th, 1987. He received a
Bachelor's degree in Mathematics from
Ege University and started an integrated

REFERENCES
[1]

PhD programme in the field of algebraic

M. Allili, K. Mischaikow, A. Tannenbaum, Cubical homology

topology at the same university. He is

and the topological classification of 2D and 3D imagery, IEEE

interested in algebraic topology and digital

International Conference on Image Processing 2(2001),

topology.

173176.
[2]
[3]
[4]

L. Boxer, Digitally continuous functions, Pattern Recognition

Letters 15 (1994), 833-839.
L. Boxer, A classical construction for the digital
fundamentalgroup, J. Math. Imaging Vis. 10 (1999), 51-62.
L. Boxer, I. Karaca, and A. Oztel, Topological Invariants in
Digital Images, Journal of Mathematical Sciences: Advances
and Applications 11(2), 2011, 109-140.
IJICS Volume 1, Issue 7 October 2012 PP. 178-187 www.iji-cs.org Science and Engineering Publishing Company
- 169 -