Digital Image Homology Analysis
Digital Image Homology Analysis
http://dx.doi.org/10.12785/amis/080529
Abstract: In this paper we are interested in relative homology groups of digital images. Some properties of the Euler characteristics for
digital images are given. We also present reduced homology groups for digital images. The main purpose is to obtain some differences
between notions in digital topology and algebraic topology.
homology groups. In the next section we introduce the is (κ0 , κ1 )-continuous and bijective and also f −1 : Y −→ X
relative homology groups for digital images, give sample is (κ1 , κ0 )-continuous. For a digital image (X, κ ) and its
calculations and provide some important properties. In subset (A, κ ), we call (X, A) a digital image pair with κ -
the section 4, we deal with properties of the Euler adjacency.
characteristics. In the last section we make some Definition 2.2. [3]. Let X ∈ Zn0 and Y ∈ Zn1 be digital
conclusions. images with κ0 -adjacency and κ1 -adjacency respectively.
For two (κ0 , κ1 )-continuous functions f , g : X −→ Y , if
there is a positive integer m and a function
2 Preliminaries H : X × [0, m]Z −→ Y such that
A pair (X, κ ) is called a digital image, where Z is the set • for all x ∈ X, H(x, 0) = f (x) and H(x, m) = g(x);
of integers, X ⊂ Zn for some positive integer n and κ
represents certain adjacency relation for the members of • for all x ∈ X, the induced function Hx : [0, m]Z −→ Y
X. We give an adjacency relation which is used in the defined by
study of digital images. Let l, n be positive integers such
that 1 ≤ l ≤ n and two distinct points Hx (t) = H(x,t) for all t ∈ [0, m]Z ,
p = (p1 , p2 , . . . , pn ), q = (q1 , q2 , . . . , qn ) is (2, κ1 )-continuous; and
in Zn . If there are at most l distinct coordinates j for
• for all t ∈ [0, m]Z , the induced function Ht : X −→ Y
which |p j − q j | = 1 and for all other coordinates j,
defined by
p j = q j then we say that p and q are kl -adjacent [5]. The
number of points q ∈ Zn that are adjacent to a given point
Ht (x) = H(x,t) for all x ∈ X,
p ∈ Zn are represented by a kl -adjacency relation. In Z,
k1 -adjacent points are called 2-adjacent; in Z2 , is (κ0 , κ1 )-continuous, they are said to be digitally
k1 -adjacent points and k2 -adjacent points are called (κ0 , κ1 )-homotopic in Y and this is denoted by
4-adjacent and 8-adjacent; finally in Z3 , k1 , k2 and f ≃(κ0 ,κ1 ) g. The function H is called a digital
k3 -adjacent points are called 6-adjacent, 18-adjacent and
(κ0 , κ1 )-homotopy between f and g. The digital
26-adjacent, respectively.
homotopy relation is equivalence relation among digitally
Let κ be an adjacency relation defined on Zn . A
continuous functions (see [3]).
κ -neighbor of p ∈ Zn is a point of Zn that is κ -adjacent to
p. A digital image X ⊂ Zn is κ -connected [10] if and only Let f : X −→ Y and g : Y −→ X be (κ0 , κ1 ) and (κ1 , κ0 )
if for every pair of different points x, y ∈ X, there is a set continuous functions respectively such that
{x0 , x1 , . . . , xr } of points of a digital image X such that
f ◦ g ≃(κ1 ,κ1 ) 1Y and g ◦ f ≃(κ0 ,κ0 ) 1X
x = x0 , y = xr and xi and xi+1 are κ -neighbors where
i = 0, 1, . . . , r − 1. A κ -component of a digital image X is
a maximal κ -connected subset of X. A set of the form We say that X and Y have the same (κ0 , κ1 )-homotopy type
[a, b]Z = {z ∈ Z|a ≤ z ≤ b} is a digital interval [2], where and that X and Y are (κ0 , κ1 )-homotopy equivalent [3].
a, b ∈ Z with a < b. Definition 2.3. [3]. (i) A digital image (X, κ ) is said to be
Let (X, κ0 ) ⊂ Zn0 and (Y, κ1 ) ⊂ Zn1 be digital images. κ -contractible if its identity map is (κ , κ )-homotopic to a
A function f : X −→ Y is (κ0 , κ1 )-continuous [3], if for constant function c̄ for some c ∈ X where the constant
every κ0 -connected subset U of X, f (U) is a function c̄ : X −→ X is defined by c̄(x) = c for all x ∈ X.
κ1 -connected subset of Y . In [3], it is shown that a
function f : X −→ Y is (κ0 , κ1 )-continuous if and only if (ii) Let (X, A) be a digital image pair with
for every κ0 -adjacent points {x0 , x1 } of X, either κ -adjacency and i : A −→ X be the inclusion function. A
f (x0 ) = f (x1 ) or f (x0 ) and f (x1 ) are a κ1 -adjacent in Y . is called a κ -retract of X if and only if there is a
In a digital image (X, κ ), a (2, κ )-continuous function κ -continuous function r : X −→ A such that r(a) = a for
f : [0, m]Z −→ X such that f (0) = x and f (m) = y is called all a ∈ A. Then the function r is called a κ -retraction of X
a digital κ -path [6] from x to y. If f (0) = f (m) then f is onto A.
a digital κ -loop and the point f (0) is the base point of the
loop f . (iii) A digital homotopy H : X × [0, m]Z −→ X is a
Definition 2.1. [4]. A simple closed κ -curve of m ≥ 4 deformation κ -retract if the induced map H(−, 0) is the
points in a digital image X is a sequence identity map 1X and the induced map H(−, m) is
{ f (0), f (1), . . . , f (m − 1)} of images of the κ -path retraction of X onto H(X × {m}) ⊂ X. The set
f : [0, m − 1]Z −→ X such that f (i) and f ( j) are H(X × {m}) is called a deformation κ -retract of X.
κ -adjacent if and only if j = i ± mod m. Definition 2.4. [12]. Let (X, κ ) be a digital image and S
Let (X, κ0 ) ⊂ Zn0 and (Y, κ1 ) ⊂ Zn1 be digital images. be a set of nonempty subset X. Assume that the following
A function f : X −→ Y is a (κ0 , κ1 )-isomorphism [6] if f hold:
c 2014 NSP
Natural Sciences Publishing Cor.
Appl. Math. Inf. Sci. 8, No. 5, 2337-2345 (2014) / www.naturalspublishing.com/Journals.asp 2339
a) If p and q are distinct points of s ∈ S, then p and q Theorem 2.8. [7]. For each q ≥ 0, Hqκ is a covariant
are κ -adjacent. functor from the category of digital simplicial complexes
b) If s ∈ S and 0/ 6= t ⊂ s, then t ∈ S. and simplicial maps to the category of abelian groups.
Corollary 2.9. [1]. If f : (X, κ1 ) −→ (Y, κ2 ) is a digitally
Then the members of S are called simplexes of (X, κ ). (κ1 , κ2 )-isomorphism, then f∗ : Hqκ1 (X) −→ Hqκ2 (Y ) is a
An m-simplex is a simplex S such that |S| = m + 1. group isomorphism.
′
Let P be a digital m-simplex. If P is a nonempty
′
proper subset of P, then P is called a face of P. Vert(P) is
used to denote the vertex set of P. A digital subcomplex A 3 Digital relative homology groups
of a digital simplicial complex X with κ -adjacency is a
digital simplicial complex [12] contained in X with Let (A, κ ) be a digital subcomplex of the digital simplicial
Vert(A) ⊂ Vert(X). complex (X, κ ). Then the chain group Cqκ (A) is a subgroup
Let (X, κ ) be a finite collection of digital m-simplices, of the chain group Cqκ (X). The quotient group
0 ≤ m ≤ d for some non-negative integer d. If the
following statements hold then (X, κ ) is called a finite Cqκ (X, A) = Cqκ (X)/Cqκ (A)
digital simplicial complex [1]:
• If P belongs to X, then every face of P also belongs is called the group of relative chains of X modulo A. The
to X. boundary operator
• If P, Q ∈ X, then P ∩ Q is either empty or a common
face of P and Q. ∂q : Cqκ (A) −→ Cq−1
κ
(A)
The dimension of a digital simplicial complex X is the
largest integer m such that X has an m-simplex. Cqκ (X) is a is the restriction of the boundary operator on Cqκ (X). It
free abelian group [1] with basis all digital (κ , q)-simplices induces a homomorphism
in X.
Cqκ (X, A) −→ Cq−1
κ
(X, A)
Let (X, κ ) ⊂ Zn be a digital simplicial complex of
dimension m. Boxer et.al. show that (see [7]) for all of the relative chain groups and this is also denoted by ∂q .
q > m, Cqκ (X) is a trivial group. The homomorphism
Definition 3.1. [9]. Let (A, κ ) be a digital subcomplex of
∂q : Cqκ (X) −→ Cq−1
κ (X) defined (see [1]) by
the digital simplicial complex (X, κ ).
q • Zqκ (X, A) = Ker ∂q is called the group of digital
∂q (u) = ∑i=0 (−1)i < p0 , p1 , . . . , p̂i , . . . , pq >, q ≤ m relative simplicial q-cycles.
0, q>m • Bκq (X, A) = Im ∂q+1 is called the group of digital
relative simplicial q-boundaries.
is called a boundary homomorphism, where p̂i means • Hqκ (X, A) = Zqκ (X, A)/Bκq (X, A) is called the qth
delete the point pi and u =< p0 , p1 , . . . , pq >. In [1], digital relative simplicial homology group.
Havana et.al. show that for all 1 ≤ q ≤ m,
Example 3.2. Let X = [0, 1]Z be the digital interval and A
∂q−1 ◦ ∂q = 0. be a single point digital image {0}. Let’s determine
Hq2 (X, A) for all q ≥ 0. For q = 0,
Definition 2.5. [1]. Let (X, κ ) be a digital simplicial
= Z and C02 (X, A) ∼
= Z ⊕ Z, C02 (A) ∼
C02 (X) ∼ = Z.
complex.
(1) Zqκ (X) = Ker ∂q is called the group of digital For q = 1,
simplicial q-cycles.
(2) Bκq (X) = Im ∂q+1 is called the group of digital C12 (X) ∼
= Z, C12 (A) ∼
= 0 and C12 (X, A) ∼
= Z.
simplicial q-boundaries.
(3) Hqκ (X) = Zqκ (X)/Bκq (X) is called the qth digital Since there are no digital (2, q)-simplexes of dimension 2
simplicial homology group. or higher, then
Theorem 2.6. [7]. Let (X, κ ) be a directed digital Cq2 (X, A) = 0
simplicial complex of dimension m. and accordingly Hq2 (X, A) = 0 for all q ≥ 2.
• Hqκ (X) is a finitely generated abelian group for every
q ≥ 0. Now consider the following short sequence:
• Hqκ (X) is a trivial group for all q > m.
• Hmκ (X) is a free abelian group, possible zero. ∂ ∂ ∂
0 →2 C12 (X, A) →1 C02 (X, A) →0 0.
Theorem 2.7. [1]. If (X, κ ) is a single point digital image,
then There are obvious results such that
κ Z, q = 0,
Hq = Im ∂2 = 0 and Ker ∂0 ∼
= Z.
0, q 6= 0.
c 2014 NSP
Natural Sciences Publishing Cor.
2340 O. Ege et. al. : Relative Homology Groups of Digital Images
The elements of C12 (X, A) are of the form Proof. Consider the following long exact sequence:
Z/Ker ∂1 ∼
= Im ∂1 ∼
= Z ⇒ Ker ∂1 = 0. C1κ (A) = {α < u j >: α ∈ Z, j ∈ J}.
is exact, where i : A −→ X and p : X −→ (X, A) are Im ∂1 = C0κ (X, A). Therefore, we have H0κ (X, A) = 0.
inclusion maps. This is called exactness axiom. Proposition 3.6. If A = {x0 } is a single point digital image,
We now give some significant results about the relative Hqκ (X, A) ∼
= Hqκ (X) for q > 0.
simplicial homology groups of digital images.
Proof. From Theorem 2.7, we know that Hqκ (A) = 0 for
Proposition 3.3. Let (X, κ ) be a digital image and A ⊂ X. q > 0. By the long sequence
If A is empty set,
∂
Hqκ (X, A) ∼
= Hqκ (X) . . . → Hqκ (A) → Hqκ (X) → Hqκ (X, A) →∗ Hq−1
κ
(A) → . . . ,
c 2014 NSP
Natural Sciences Publishing Cor.
Appl. Math. Inf. Sci. 8, No. 5, 2337-2345 (2014) / www.naturalspublishing.com/Journals.asp 2341
Therefore i∗ : Hqκ (A) −→ Hqκ (X) is an (κ , κ )-isomorphism ∂∗ is the trivial map because Ker i∗ = 0. Thus j∗ is an
for all q ≥ 0. From the following long exact sequence κ (A), we have
epimorphism. Since Ker j∗ = Im i∗ = Hq−1
∂ κ κ κ
. . . → Hqκ (A) → Hqκ (X) → Hqκ (X, A) →∗ Hq−1
κ
i
∗ j∗
(A) → . . . , Hq−1 (X)/Ker j∗ = Hq−1 (X)/Hq−1 (A)
∼ κ
= H (X, A)
j∗ : Hqκ (X) −→ Hqκ (X, A) must be a zero map. By q−1
exactness, ∼
= Im j∗
Hq8 (X, A) = 0.
′ ′
So we have the following short sequence: MSC8 is digital 8-retract of MSC4 . Since MSC8 and MSC4
are digital simple closed 8-curve,
i j k
0 → H18 (X, A) → Z → Z → H08 (X, A) → 0.
8 8 ′ Z, q = 0, 1
Hq (MSC4 ) = Hq (MSC8 ) =
By Theorem 3.5, we get H08 (X, A) = 0. From the First 0, q 6= 0, 1
Isomorphism Theorem and the fact that Ker j = Im i = 0,
(see [1]). Consider the following long exact homology
H18 (X, A)/Ker j∼ = Ker k ∼
= Im j ∼ = 0 ⇒ H18 (X, A) = 0. sequence:
′ ′
As a result, we have Hq8 (X, A) = 0 for all q ≥ 0. · · · → H28 (MSC8 ) → H28 (MSC4 ) → H28 (MSC4 , MSC8 )
Theorem 3.10. Let (X, A) be a digital image pair with κ -
adjacency such that (A, κ ) is a κ -retract of (X, κ ). Then ∂ ∗ ′ i j∗ ′
→ H18 (MSC8 ) → H18 (MSC4 ) → H18 (MSC4 , MSC8 )
Hqκ (X) ∼
= Hqκ (A) ⊕ Hqκ (X, A). ∂ ∗ ′ i j∗ ′
→ H08 (MSC8 ) → H08 (MSC4 ) → H08 (MSC4 , MSC8 ) → 0
Proof. Since (A, κ ) is a κ -retract of (X, κ ), r ◦ i = 1(A,κ ) So we get
where i : (A, κ ) −→ (X, κ ) and j : (X, 0)
/ −→ (X, A) are
∂
inclusion maps and r : (X, κ ) −→ (A, κ ) is a κ -retraction j∗ j∗
∗ i ∗ ′ i
0→Z→ Z → H18 (MSC4 , MSC8 ) → Z → Z→0
map. So r∗ ◦ i∗ = 1Hqκ (A) and i∗ is a monomorphism.
′
Consider the exactness of the homology sequence of By this exact sequence we have H18 (MSC4 , MSC8 ) = 0. As
(X, A) a result, for all q ≥ 0
∂ ∂
. . . → Hqκ (X, A) →∗ Hq−1
κ κ κ
i j∗ ′ ′
∗
(A) → Hq−1 (X) → Hq−1 (X, A) →∗ . . . Hq8 (MSC4 ) ∼
= Hq8 (MSC8 ) ⊕ Hq8 (MSC4 , MSC8 ).
c 2014 NSP
Natural Sciences Publishing Cor.
2342 O. Ege et. al. : Relative Homology Groups of Digital Images
We now calculate the relative homology groups of • Using the exact sequence, we get
′
(MSS6 , A) where ′
H04 (MSS6 , A)/Ker δ1 ∼
= Im δ1 = Ker k∗ = 0
′
A = {c0 , c1 , c2 , c3 } ⊂ MSS6 ′
⇒ H04 (MSS6 , A) = Ker δ1 = Im j∗ .
has 6-adjacency.
′ Similarly, since Ker j∗ = Im i∗ = Z5 we have
Example 3.12. Let MSS6 = {c0 = (0, 0, 0), c1 =
(1, 0, 0), c2 = (1, 1, 0), c3 = (0, 1, 0), c4 = (0, 0, 1), c5 = Z5 /Ker j∗ ∼
= Im j∗ ⇒ Im j∗ = 0
(1, 0, 1), c6 = (1, 1, 1), c7 = (0, 1, 1)} (see Figure 2).
′
and consequently H14 (MSS6 , A) = 0.
′
As a result, for all q ≥ 0, we have Hq4 (MSS6 , A) = 0.
4 ′ ∗ δ i ′
· · · → Hq+1 (MSS6 , A) → Hq4 (A) → Hq6 (MSS6 )
j∗ ′
Define a map H : ZP1 × [0, 2]Z −→ ZP1 by the following.
→ Hq4 (MSS6 , A) → . . .
H(c, 0) = c, t =0
Then we have
H(c7 , 1) = c0 , H(c2 , 1) = c1 , t = 1
p∗ ′ ∗ δ i ∗ j∗ ′ δ k
H(c , 2) = c , t =2
0 → H24 (MSS6 , A) →2 Z → Z5 → H14 (MSS6 , A) →1 Z → Z 1 0
c 2014 NSP
Natural Sciences Publishing Cor.
Appl. Math. Inf. Sci. 8, No. 5, 2337-2345 (2014) / www.naturalspublishing.com/Journals.asp 2343
Consider the following long exact homology sequence: Proof. Since the sequence
ε
∗
· · · → Hq4 (ZP1 ) →
i j∗
Hq4 (MSC4 ) → Hq4 (MSC4 , ZP1 ) 0 → Ker ε → C0κ (X) → Z → 0
δ is exact,
4
→ Hq−1 (ZP1 ) → . . .
Ker ε Cκ (X) ε
Then we have 0→ → 0 →Z→0
Im ∂1 Im ∂1
i p ∂ i
2
0→ Z →2 H14 (MSC4 , ZP1 ) →1 Z →
1
Z is also exact. It follows from Z is free that the above
sequence splits and we find that
p ∂
→1 H04 (MSC4 , ZP1 ) →0 0.
H0κ (X) ∼
=Hfκ (X) ⊕ Z.
0
By this exact sequence we have some conclusions:
• Since Im i2 = 0 = Ker p2 and from the First Isomorphism The other statement is a trivial result.
Theorem, we have
If we make a little change the digital simplicial They also prove that if (X, κ ) is a digital image of
homology, we get a new concept which is reduced digital dimension m, then
homology. The only difference between the two m
homologies lies on the group H0κ (X). χ (X, κ ) = ∑ (−1)q rankHqκ (X).
Definition 3.14. [12]. Let (X, κ ) be a digital simplicial q=0
complex. We define a homomorphism ε : C0κ (X) −→ Z by
putting ε (v) = 1 for each vertex v of X and extending We describe a few basic properties of the Euler
linearly to C0κ (X). This homomorphism, called characteristics. The Euler characteristics is a function χ
augmentation map, is onto. If ∂1 : C1κ (X) −→ C0κ (X) is which associates to each digital images (X, κ ) an integer
the boundary homomorphism, then ε ◦ ∂1 = 0. χ (X, κ ). The Euler characteristics is determined by the
following properties which holds in Algebraic Topology.
fκ (X) = Ker ε
Let X and Y be any topological spaces, and {p} be a
H single point set.
0
Im ∂1
1.χ ({p}) = 1.
is called the zero dimensional reduced digital homology 2.If X is homeomorphic to Y then χ (X) = χ (Y ).
fpκ (X) = Hpκ (X) for each p > 1,
group of (X, κ ). If we put H 3.For any homotopic compact spaces X and Y ,
then
fκ (X), i = 0, 1, . . .}
{H χ (X) = χ (Y ).
i
are called the reduced digital homology groups of (X, κ ). 4.For every closed subset C ⊂ X,
There is an induced inclusion map H fκ (X) −→ H κ (X)
0 0
because Im ∂1 ⊂ Ker ε ⊂ C0κ (X). χ (X) = χ (C) + χ (X \C).
Theorem 3.15. [12]. For a digital simplicial complex
(X, κ ), there are the following formulas which are related This property has a dual form
to reduced homology groups :
χ (X) = χ (U) + χ (X \U),
H0κ (X) ∼
=Hfκ (X) ⊕ Z,
0 for every open subset U ⊂ X.
5.χ (X ×Y ) = χ (X).χ (Y ).
Hpκ (X) = H
fpκ (X), p ≥ 1.
c 2014 NSP
Natural Sciences Publishing Cor.
2344 O. Ege et. al. : Relative Homology Groups of Digital Images
χ (X ×Y ) 6= χ (X).χ (Y ).
Fig. 4: X \C
5 Conclusion
Example 4.2. Let X be the digital image MSC4 . We take
the digital projective line ZP1 which is the closed subset The main goal of this study is to compute the relative
of X. Let this image be denoted by C. Consider a digital homology groups and determine some properties of Euler
image X \ C. This is again the digital projective line ZP1 characteristics for digital images. We hope that this article
(see Figure 4). The Euler characteristics of X, C and X \C will be helpful in the study of digital homology groups
are and that will yield us to compute the homology groups of
some part of the digital image when we know the
χ (X) = 0, χ (C) = 1 and χ (X \C) = 1, homology groups of an entire image.
c 2014 NSP
Natural Sciences Publishing Cor.
Appl. Math. Inf. Sci. 8, No. 5, 2337-2345 (2014) / www.naturalspublishing.com/Journals.asp 2345
c 2014 NSP
Natural Sciences Publishing Cor.