Download as pdf or txt
Download as pdf or txt
You are on page 1of 45

Fractal Descriptors Applied to Image Analysis

Joao B. Florindo
Physics Institute of Sao Carlos - University of Sao Paulo - Brazil
College of Medical and Dental Sciences - University of Birmingham
jbflorindo@gmail.com
http://scg.ifsc.usp.br

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

1 / 45

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

2 / 45

University of Sao Paulo


Top university in Latin-America (QS World University Rankings)
Numbers (approximate):
Established as university in 1934
93,000 students (30,000 PhD)
Total area: 76 million m2

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

3 / 45

Introduction

Fractal Geometry

Fractal Descriptors

Results

Other Works

Conclusions

Bibliography

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

4 / 45

Introduction

Introduction

Fractals are geometrical objects defined within an infinite range of


scales
Complex and self-similar structures at any scale

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

5 / 45

Introduction

Introduction

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

6 / 45

Introduction

Introduction
Nature shows many fractal-like processes [Mandelbrot, 1982]
Complex and self-similar objects (within a limited range of scales)

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

7 / 45

Introduction

Introduction
Fractal Theory has been widely applied to image analysis
[Tian-Gang et al., 2007, Xu et al., 2009]
Good results in texture and shape analysis
[Xu et al., 2009, Florindo and Bruno, 2013a, Backes et al., 2012]

TEXTURES

Joao B. Florindo

SHAPES

Fractal Descriptors Applied to Image Analysis

8 / 45

Introduction

Introduction

The most important measure of a fractal is its


dimension[Mandelbrot, 1982]
This can be adapted to describe objects in nature
Limited information / scale-dependent
Multifractals, multiscale fractal dimension, fractal descriptors

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

9 / 45

Fractal Geometry

Fractal Geometry

Fractal dimension measures how the amount of details in the object


changes according to scale
Mathematically, it is a real non-negative number provided by the
Hausdorff-Besicovitch dimension
Such concept cannot be directly applied to finite objects (and images)
but can be approximated

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

10 / 45

Fractal Geometry

Fractal Geometry
A general expression can be defined for real-world objects
(fractal-like) and applied to the respective image representation
D lim

0

Joao B. Florindo

log(M())
log()

Fractal Descriptors Applied to Image Analysis

11 / 45

Fractal Geometry

Fractal Geometry

Fractal Geometry in Computer Vision


Euclidean Geometry is limited to describe the complexity of nature
Many objects in nature present high complexity and self-similarity and
can be modelled by fractals
Fractal dimension measures the complexity or spatial occupation

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

12 / 45

Fractal Descriptors

Fractal Descriptors

Fractal dimension alone is scale-dependent and it is not sufficient in


more complex problems
Multiscale Fractal Dimension [Bruno et al., 2008]

Figure based on [Bruno et al., 2008]

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

13 / 45

Fractal Descriptors

Fractal Descriptors

Descriptors derived from the definition of descriptors in images:


Fourier, wavelets, etc.
It uses the entire fractality curve
Descriptors
u : log() log(M).
Complete mapping of object complexity, either in microstructures or
in a global viewpoint

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

14 / 45

Fractal Descriptors

Fractal Descriptors

SPECTRUM

TEXTURE

IMAGE

SHAPE

Joao B. Florindo

COLOUR

Fractal Descriptors Applied to Image Analysis

15 / 45

Fractal Descriptors

Fractal Descriptors: Texture

SPECTRUM

TEXTURE

IMAGE

SHAPE

Joao B. Florindo

COLOUR

Fractal Descriptors Applied to Image Analysis

16 / 45

Fractal Descriptors

Fractal Descriptors: Multiscale transform


Multiscale transform [Florindo and Bruno, 2013b]
Space-scale
Z

U(b, a) =

u 0 (t b)ga dt

n=

Time-frequency
Z

U(b, a) =

g (t b)u(t)e 2at dt

Time-scale
1
U(b, a) =
a
Joao B. Florindo

Z
(
<

t b
)u(t)dt
a

Fractal Descriptors Applied to Image Analysis

17 / 45

Fractal Descriptors

Fractal Descriptors: FDA


Functional data analysis [Florindo et al., 2011]
Measures on log-log curve are addressed through an analytical
function
The descriptors are obtained from the coefficients of the projection
onto a set of basis functions
FDA
yji = u i (xji ) + ij
0

ui =

q
X

j (u i )j

j=1

(u) = S(u)
Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

18 / 45

Fractal Descriptors

Fractal Descriptors: Multilevels


Multilevel descriptors [Florindo and Bruno, 2013a]

...
1,0

1,0

0,8

0,6

D(k)

0,2

10

20

30

40

50

60

70

80

90

10

20

30

40

50

0,8
0,6

D(k)

0,8
0,6

70

80

90

60

70

80

90

0,4

0,4

0,2

0,2

20

30

40

50

60

70

80

90

10

20

30

40

50

1,0

0,8

0,8

0,6

0,6

D(k)

1,0

0,4

0,4

0,2

0,2

D(k)

1,0

1,0

1,0

0,8

0,8

0,8

0,6

10

20

30

40

50

60

70

80

D1M

90

10

20

30

40

50

D1

60

70

80

0,4

0,2

0,2

10

20

30

40

50

90

D2M

60

70

80

90

10

20

30

40

D2

50

60

70

80

90

1,0
0,8

0,6

0,4

0,2
0
0

0,6

0,4

D(k)

10

D(k)

D(k)

60

k
1,0

1,0

D(k)

0,6
0,4

0,4
0,2

D(k)

D(k)

0,8

0,6
0,4
0,2

10

20

30

40

50

60

70

80

90

10

20

30

40

50

60

70

80

90

DlmaxM Dlmax

i = [D1M(i) D2M(i) ... DlmaxM(i)]

i = [D1(i) D2(i) ... Dlmax(i)]

EFV = [K(1) K(2) ... K(n) K(1) K(2) ... K(n)]

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

19 / 45

Fractal Descriptors

Fractal Descriptors: Window-based


Used in large dimension images, as usual in nanotechnology, and
segmentation
TRAINING

TESTING

WINDOWS

WINDOWS

DESCRIPTORS

DESCRIPTORS

CLASSIFIER

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

20 / 45

Fractal Descriptors

Fractal Descriptors: Colour

SPECTRUM

TEXTURE

IMAGE

SHAPE

Joao B. Florindo

COLOUR

Fractal Descriptors Applied to Image Analysis

21 / 45

Fractal Descriptors

Fractal Descriptors: Colour


Combining descriptors from EEE channels
[Florindo and Bruno, 2011b]

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

22 / 45

Fractal Descriptors

Fractal Descriptors: Shapes

SPECTRUM

TEXTURE

IMAGE

SHAPE

Joao B. Florindo

COLOUR

Fractal Descriptors Applied to Image Analysis

23 / 45

Fractal Descriptors

Fractal Descriptors: Curvature (shapes)


Curvature fractal descriptors [Backes et al., 2012]

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

24 / 45

Fractal Descriptors

Fractal Descriptors: Spectrum

SPECTRUM

TEXTURE

IMAGE

SHAPE

Joao B. Florindo

COLOUR

Fractal Descriptors Applied to Image Analysis

25 / 45

Fractal Descriptors

Fractal Descriptors: Gabor


Gabor-transformed fractal descriptors [Zuniga et al., 2014]

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

26 / 45

Results

Results: Brodatz

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

27 / 45

Results

Results: Brodatz
1

T1

0.995

D(k)

0.99

0.985

T2

0.98

0.975

0.97

Joao B. Florindo

T1
T2
0

10

20

30

Fractal Descriptors Applied to Image Analysis

40

50

60

28 / 45

Results

Results: Brodatz

KNN

Joao B. Florindo

Method
Multifractals
LBP
Fourier
GLCM
Gabor

ND
17
20
16
20
7

CR (%)
92.61
92.61
89.01
88.83
86.85

Bouligand-Minkowski
Multilevels + TS
Probability + TS
Lacunarity + TS
Wavelets + TS
Triangular + TF
Fourier

14
18
8
10
8
10
10

99.10
98.92
92.25
87.75
84.14
82.52
82.16

Lacunarity+Multilevels
Brownian+Multilevels
Fourier+Multilevels
Probability+Multilevels + TS
Bouligand-Minkowski+Multilevels + TS
Triangular+Multilevels
Lacunarity+Wavelets + TS

11
19
20
16
20
19
13

99.46
99.46
99.46
99.28
99.28
99.10
94.95

Fractal Descriptors Applied to Image Analysis

29 / 45

Results

Results: Taxonomy [Florindo et al., 2014]

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

30 / 45

Results

Results: Taxonomy [Florindo et al., 2010]

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

31 / 45

Results

Results: Microscopy [Silva et al., 2014]

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

32 / 45

Results

Results: Nanotechnology [Florindo et al., 2013]

0.4

D(r)

0.2

-0.2

-0.4
0

Joao B. Florindo

20

40

60

Fractal Descriptors Applied to Image Analysis

33 / 45

Results

Results: Segmentation

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

34 / 45

Other Works

Other Complexity-Based Methods

Fractal Geometry is an example of complexity modelling


Here complexity means the degree of heterogeneity of a texture or
irregularity of a shape
Other examples of complexity-based methods are Complex Networks,
Deterministic Partially Self-avoiding Walks, etc.
Complex Networks and Fractal Geometry are related
[Song et al., 2005].

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

35 / 45

Other Works

Other Complexity-Based Methods


DIFFUSION DYNAMIC

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

36 / 45

Other Works

Other Complexity-Based Methods

BOX-COVERING

log NB (lB )
lB 0
log lB

D lim

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

37 / 45

Other Works

Other Complexity-Based Methods

CLUSTERING COEFFICIENT

D lim

d0

Joao B. Florindo

log k(d)
log d

Fractal Descriptors Applied to Image Analysis

38 / 45

Conclusions

Conclusions

Fractal descriptors express all the complexity of the object


They capture nuances at different levels of scale
They discriminate among complex materials automatically and with
high precision
Inexpensive technique (both in terms of computational and money
resources) and providing promising results in different applications

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

39 / 45

Conclusions

Thank you all!


Questions ...

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

40 / 45

Bibliography

Bibliography I

Backes, A. R., Florindo, J. B., and Bruno, O. M. (2012).


Shape analysis using fractal dimension: a curvature based approach.
Chaos, 22(1):043103.
Bruno, O. M., de Oliveira Plotze, R., Falvo, M., and de Castro, M.
(2008).
Fractal dimension applied to plant identification.
Inf. Sci, 178(12):27222733.
Florindo, J. and Bruno, O. (2013a).
Texture analysis by multi-resolution fractal descriptors.
Expert Systems with Applications, 40(10):40224028.

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

41 / 45

Bibliography

Bibliography II
Florindo, J. B., Backes, A. R., and Bruno, O. M. (2010).
Leaves shape classification using curvature and fractal dimension.
In IMAGE AND SIGNAL PROCESSING, PROCEEDINGS, volume
6134 of Lecture Notes in Computer Science, pages 456462,
HEIDELBERGER PLATZ 3, D-14197 BERLIN, GERMANY.
SPRINGER-VERLAG BERLIN.
Florindo, J. B. and Bruno, O. M. (2011a).
Closed contour fractal dimension estimation by the fourier transform.
Chaos Solitons & Fractals, 44(10):851861.
Florindo, J. B. and Bruno, O. M. (2011b).
Fractal descriptors in the fourier domain applied to color texture
analysis.
Chaos, 21(4):110.
Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

42 / 45

Bibliography

Bibliography III
Florindo, J. B. and Bruno, O. M. (2013b).
Multiscale fractal descriptors applied to texture classification.
Journal of Physics: Conference Series, 410(1):012022.
Florindo, J. B., da Silva, N. R., Romualdo, L. M., de Fatima da Silva,
F., de Cerqueira Luz, P. H., Herling, V. R., and Bruno, O. M. (2014).
Brachiaria species identification using imaging techniques based on
fractal descriptors.
Computers and Electronics in Agriculture, 103(0):48 54.
Florindo, J. B., De Castro, M., and Bruno, O. M. (2011).
Enhancing volumetric bouligand-minkowski fractal descriptors by using
functional data analysis.
International Journal of Modern Physics C, 22(9):929952.

Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

43 / 45

Bibliography

Bibliography IV
Florindo, J. B., Sikora, M. S., Pereira, E. C., and Bruno, O. M.
(2013).
Characterization of nanostructured material images using fractal
descriptors.
Physica A: Statistical Mechanics and its Applications, 392(7):1694
1701.
Mandelbrot, B. B. (1982).
The Fractal Geometry of Nature.
Freeman.
Silva, N. R., Florindo, J. B., G
omez, M. C., Kolb, R. M., and Bruno,
O. M. (2014).
Fractal descriptors for discrimination of microscopy images of plant
leaves.
Journal of Physics: Conference Series, 490(1):012085.
Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

44 / 45

Bibliography

Bibliography V
Song, C., Havlin, S., and Makse, H. A. (2005).
Self-similarity of complex networks.
Nature, 433(7024):392395.
Tian-Gang, L., Wa, S., and Zhao, N. (2007).
Fractal research of pathological tissue images.
Computerized Medical Imaging and Graphics, 31(8):665671.
Xu, Y., Ji, H., and Ferm
uller, C. (2009).
Viewpoint invariant texture description using fractal analysis.
Int. J. Comput. Vision, 83(1):85100.
Zuniga, A. G., Florindo, J. B., and Bruno, O. M. (2014).
Gabor wavelets combined with volumetric fractal dimension applied to
texture analysis.
Pattern Recognition Letters, 36:135143.
Joao B. Florindo

Fractal Descriptors Applied to Image Analysis

45 / 45

You might also like