SIZE AND SHAPE ANALYSIS OF MINERAL PARTICLES USING IMAGE PROCESSING

TECHNIQUE
D. Karaku, A. H. Onur, A.H. Deliormanl, G. Konak

Mining Engineering Department, Faculty of Engineering, Dokuz Eylul University, Izmir, Turkey

ABSTRACT: Detection of size and shape of minerals are important for the information about minerals.
Different size and shape of same mineral can be effective in classification of them. Therefore it is necessary to
know the size and shape of mineral exactly. In this study, using possibilities of image processing techniques in the
detection of shape and size of mineral was discussed and a sample study was carried out. According to this,
successful results can be obtained about size and shape features of minerals by using image processing techniques
especially in thin section images whose third dimension effect was decreased.

Keywords: Image processing, thin section, minerals size, minerals shape

1. INTRODUCTION
With the development of technology image taking
technology has also developed and image processing
methods have taken their place in many disciplines as
standard measurement and evaluation method. Fast
taking of images, detection of color and physical size
over image, advantage of analysis in short time made
image
processing
methods
superior
over
conventional analysis methods. Parallel to this, being
adherent to common image processing terminology
studies were carried out in different disciplines.
Especially in the fields in which measurement is
difficult such as physical and engineering
characteristics of rocks, researchers evaluated image
analysis methods alternatively. Studies were carried
out under microscope on the identification of
minerals, detection of void ratio (porosity) of rocks,
analysis of discontinuity, detection of mineral
boundary, detection of grain boundary, grain size
distribution and formal analysis.
With the development of image taking technology
with electron microscope and polarized microscopes
studies about identification of minerals, detection of
texture characteristics were carried out.

micro sized researches were carried out and spaces

were detected on the images taken from optic and
electron microscope. Studies on images taken by
electron microscope from the surface of polished
rock was observed by Bodziony et al., (1993) and
Towey and Huslow (1995), this method was used on
many rocks and surfaces (4,5). Meng (1996) have
done analysis between spaces in marbles and fractal
size by using image processing techniques (6). With
these studies approaches have been carried out on
the gas absorption of marbles.
The analysis of micro cracks on rocks with image
processing techniques was carried out with the
principle in porosity studies. Cracks on rocks can be
seen under microscope and images can be taken.
Especially filled and empty cracks are seen as black
and analysis can be carried out. Studies with this
technique was carried out by Zheng (1989), especially
development of micro cracks on rocks under load
was detected by image processing techniques (7).
Although these studies were limited by the twodimensional analysis of three-dimensional cracks; this
problem was solved with the application of streonet
models.

In their study carried out in 1991, Clelland and Fens

did element assignment with the images taken from
electron microscope in a short time(1). Tovey and
Krinsley (1991) searched the analysis and
classification of structural analysis of minerals by
developing same methods (2). Later on, image
analysis methods were especially used in quality
control of industrial minerals. For example, the
relationship of potassium control with the texture
and degree of freedom of minerals in the production
of iron was observed by Stirling (1989) (3).

With image processing, studies on detection of grain

boundary were carried out. In order to do this, image
passes through preprocessing and segmentation
processes and grain boundaries become evident. In
their study (1998) Goodchild and Fueten expressed
and analyzed grain boundaries of grained rock image
taken from electron microscope in line by using
image processing techniques (8). With the same
principle, Jenkins et al., (1991) studied industrial
practice of image processing method, and studied the
effect of grain sizes in refractor magnesite
production (9).

Image processing techniques were also used in the

detection of void ratio of rocks. In these studies

In the studies about grain size distribution, twodimensional effect of image was tried to be resolved

Volume 12 - Issue 23

with high-resolution images. It was expressed that

even in this condition grain shape has effect on grain
size distribution. Comparison of grain size
distribution with image processing techniques with
sieve analysis values was carried out by Friedman for
the first time (1962). In his study Friedman did
statistical studies in sandstone thin sections and
obtained high correlation coefficients. Particle-size
distribution found a utilization place especially in
aggregate production (10).

an image is a variable that changes continuously both

as level and as location (18). Detection of boundaries
of different substances and objects of luminance
values in image as pixels in computer enabled
usability of image processing methods. In this sense
producing data with analysis of images obtained by
imaging devices especially in the field of medicine has
been pioneer studies and later on found a utilization
field for itself being a fast and easy measurement
technique in many disciplines.

Persson (1996), Wang (1997), Maerz (1998), Kemeny

et al. (2002) carried out studies on size distribution
after detonation and grain distribution on carriage
band (11,12,13,14).

Application of image analysis methods in geology

were carried out for the first time in the field of
measurement and identification of spaces. Parallel to
this, analysis of discontinuity and grain sizes on rocks
were carried out by using different features of image
that can be identified such as color.

In the science and engineering of geology, mineral

particles are important elements for analysis. To
obtain quantitative information about particle size
and shape from mineral, it is common that optical
microscopy and image technique were used in the
characterization of the surface roughness and
microstructure of the minerals (15, 16, 17).
1.1 Image Processing
Image is a two-dimensional sign record. It can be
either visible like photograph or a record on
magnetic band or numerical values in computer
memory. Images can be continuous-discrete, analognumerical or continuous-numerical. Luminance on

In addition to many commercial programs that

perform numerical image analysis, processes of the
analysis common in all of them are given by and
large in Figure 1.
2. MATERIAL AND METHODS
2.1 Material
In the study, size and shape analysis of minerals were
carried out with image analysis method. Many studies
were carried out on thin section images for
petrographic description.

Figure 1. General flow chart of image processing (19)

Examples involve quantifying quartz micro crack

patterns to analyze stone deterioration, conducting
porosimetric studies of stone, examining grain-size
distribution
and
grain
shape
in
stone,
comprehensively
evaluating
deterioration
mechanisms of stone, studying conservation
treatments of stone, characterizing cementitious
materials, analyzing ceramic materials, and measuring
layer thickness on a variety of materials. Background
issues that must be understood before undertaking or
evaluating image analysis include experimental design

and statistical validity, calibration, image quality, and

image enhancement (preprocessing of images). Early
work often required in-house computer programs;
recent studies make use of new comprehensive image
analysis packages. With the emergence of these packages that provide fast preprocessing and
measurement of a wide range parameters (20)
Here, application on thin section with image analysis
method was presented and a sample was carried out

D. Karaku, A. H. Onur, A.H. Deliormanl, G. Konak / The Journal of ORE DRESSING 2010

about the detection of size and shape analysis of

minerals.
2.2. Experimental Methods
Thin section image to be used was taken and
prepared from a marble sample. Image enrichment
processes were done on thin section image first of all
and image was edged.
In this process, contrast enrichment and histogram
equalization treatments were applied.
Linear Contrast Enhancement, luminance values
between 0-255 in grey level image is narrow in some
of the images. For example minimum luminance
value of an image can be 50 and maximum 100. In
such condition, recognizing the details in the image is
difficult. By contrast enhancement, luminance value
of the image is withdrawn between 0-255 and point
of image is enriched.
Contrast improving is about how far does the object
histogram to be displayed is widened. In other words
it is the minimum and maximum size of ranges to be
displayed. While this range is better displayed, values
above and below this range would not be displayed.
In linear contrast improving method, a linear
relationship is carried out between original image
values and values displayed. Linear contrast
improving gives the best result in Gauss distribution
of image histogram. An ideal gauss histogram is a
histogram in which values are places in single-mod
range and mod is distributed symmetrically at both
sides. If there are materials which give close
reflection values in the image, histogram will be gauss
or close to gauss. Values to be obtained in linear
contrast improving method are detected according to
Equation 2.

g ( x, y) (

f ( x, y) f ( x, y) min
) * 255
f ( x, y) max f ( x, y) min

g ( x, y )

: Pixel value to be displayed

f ( x, y )

: Original image pixel value

f ( x , y ) min : Original image minimum pixel value

f ( x, y) max : Original image maximum pixel value

Histogram equalization is a suitable image

enrichment method for the images whose color
values were not distributed uniform. It can either be
applied all through the image or only on a specific
part. It is called global histogram equalization if
applied all through the image and called local
histogram equalization if applied on a specific part of
image.
In histogram equalization method, equalization
algorithm is applied on original image and equal pixel
is provided per ton. With this algorithm it is ideally
expected from the histogram graphic to have linear
level that gives close values.
In histogram
equalization, first of all cumulative histogram is
found by making use of the histogram of original
image. Cumulative histogram is the graphic that
includes the values obtained by the addition of each
luminance value with the ones previous of it and
itself (Figure 2). Cumulative histogram values are
normalized by multiplying maximum luminance
value we want in enriched image and dividing the
total pixel number in the image. If normalized
histogram values and luminance values of the image
is updated we have succeeded in applying histogram
equalization method on that image.

Figure 2. Histogram equalization method

Applying contrast enhancement and histogram

equalization method on the original image, the image
is prepared for segmentation process. In this process
which is called image enrichment, boundaries in the
image are edged and the success of segmentation
process is enhanced.
In Figure 3, sample image application of contrast
enhancement and histogram equalization is given.

Volume 12 - Issue 23

segmentation produces raw information such as the

boundaries, shape and area of the object in the
image. If we are concerned about the shape of
objects, we expect segmentation to give us
information about the sides, edges and boundaries of
that object. But if we are concerned about the
internal features such as surface cover, area, colors
and frame of the object local segmentation must be
used. For the solution of very complex problems
such as identification of characteristics or generally
pattern, both segmentation methods can be used
(21). Edge detect application for the thin section
image after enrichment process is given in Figure 4.

Figure 3. Image enhancement of sample image, (a)

Original Image, (b) Contrast Enhancement, (c)
Histogram Equalization

Figure 4. Segmentation of sample image, (a)

enhancement image, (b) edge detect application of
sample image

After preprocesses are carried out segmentation

treatment begins. Segmentation is the process of
separating object and background in an image or
different parts studied in the image from themselves.
Segmentation is the most difficult application of
image processing and a specific error ratio may occur
at the end of segmentation techniques. Image

3. RESULTS AND DISCUSSIONS

3.1. Size Analysis of Mineral Particles
The most common measurement used for the sizes
of irregular grains under microscope is Ferret
diameters of grains. Ferret diameter limits can be

D. Karaku, A. H. Onur, A.H. Deliormanl, G. Konak / The Journal of ORE DRESSING 2010

defined as the farthest distance in the boundaries of a

grain. In Figure 5 ferret diameter of a grain is given.

Ferret diameters of the grain in the image are

detected. According to this, average ferret diameter
of 151 mineral grains in sample image is 808.0
micronmeter. Distribution graphics according to
ferret diameter is given in Figure 7.
25

Mean
StDev
N

23

808.0
474.7
151

20
18

18

Frequency

16

15

15
13
11

10

10

4
1

Figure 5. Definition of Ferret diameters in x direction

(Fx), y direction (Fy), and maximum Ferret diameter,
Fm (maximum dimension, Dm) (22)

As a result of image enrichment and segmentation

processes of different minerals in sample thin section
image, edges are detected. While detecting the edges,
shape features of all the grains in the image are
presented regardless of the type of minerals.
However it is possible to do optional grain analysis
according to image type by using image processing
methods.
With image analysis program, grains in the image are
assigned to the variables according to their
boundaries. The image located according to their
boundaries is given in Figure 6. Following this
process, size analysis can be carried out by making
use of each grain and located boundary of that grain
in computer memory.

Figure 6. Image located with its boundaries

300

600
900
1200
1500
1800
Feret Diameter (mikronmeter)

2100

Figure 7. Distribution graphics according to ferret

diameters
3.2. Shape Analysis of Mineral Particles

Dozens of size parameters are possible; therefore,

they can be combined in hundreds of ways into a
formally dimensionless expression that might be used
as a shape descriptor. In fact, only a few relatively
common combinations are possible, but even these
are plagued by total inconsistency in naming
conventions. Table 1 summarizes some of the most
widely used shape parameters calculated as
combinations of size measurements.
Table 1. Representative shape descriptors

Volume 12 - Issue 23

Formal features of grains described as ratio generally

aims to produce numerical data about the forms of
them by using diameter, perimeter and area
information of grains. Formal features of 151 mineral
grains described in sample image and identifies in
Ferret diameters are calculated according to the
equations given in Table 2. According to this, form
factor, roundness, compactness, aspect ratio and
solidity normal distribution graphics are given in
Figure 8.

30

Mean
StDev
N

With the formula described in Table 1, an approach

can be presented for the shape of grains in the image.
When the statistical evaluation of data as a result of
analysis is done, numerical data can be produced
about the form, roundness, compactness ratio with
average values and solution offers can be suggested
for the other mechanical features in which the shape
of grain is important. Statistical evaluation of data
obtained as a result of sample image shape analysis is
given in Table 2.

20

0.5985
0.1464
151

Mean
StDev
N

0.5826
0.1791
151

25
15

Frequency

Frequency

20

15

10

10
5
5

0.15

0.30

0.45

0.60
Form factor

0.75

0.90

0.15

0.30

0.45
0.60
Roundness

0.90

(b)

(a)
Mean
StDev
N

25

80

0.7538
0.1238
151

Mean
StDev
N

70

20

1.968
1.100
151

60

Frequency

Frequency

0.75

15

10

50
40
30
20

5
10
0

0.3

0.4

0.5

0.6
0.7
Compactness

0.8

0.9

1.0

6
8
Aspect Ratio

10

12

(d)

(c)
50

Mean
StDev
N

0.8375
0.09579
151

Frequency

40

30

20

10

0.4

0.5

0.6

0.7
Solidity

0.8

0.9

1.0

(e)
Figure 8. Normal distribution graphics: (a) form factor, (b) roundness, (c) compactness, (d) aspect ratio,
(e) solidity

D. Karaku, A. H. Onur, A.H. Deliormanl, G. Konak / The Journal of ORE DRESSING 2010
Variable
Form Factor
Roundness
Compactness
Aspect Ratio
Solidity

Table 2. Statistical evaluation of shape parameters

Total Count Mean Standard Deviation Minimum Value
151
0.59
0.15
0.16
151
0.58
0.18
0.08
151
0.75
0.12
0.29
151
1.97
1.10
1.03
151
0.84
0.09
0.34

5. CONCLUSIONS
Results obtained in this paper in which the utilization
possibilities of image processing techniques in
mineral size and shape analysis are given in articles
below:
Image processing used commonly in different
disciplines and has commercial aspect can be used
in mineral identification, size and shape analysis
as well. According to the academic studies about
the issue, mineral size and shape analysis of image
processing methods are in the way of forming and
using commercial products.
Restrictive side of image processing methods in
application is their third-dimension effect.
Therefore in this study thin section image on
which the third-dimension effect is not
determining was detailed and size and shape
analysis of grains in the image are presented.
Ferret diameters of grains in the sample image are
calculated by using image processing methods.
Formal features of grains in the same image; such
as the ratio of long diameter by short diameter,
form roundness are calculated.
Apart from the disadvantage of third-dimension
effect of image processing techniques, there is
sampling disadvantage as well. While image
processing techniques that enable twodimensional analysis in the image area are applied,
sampling size must be detected well. Size of
masses must be regarded during sample taking,
and if required numerical data must be produced
by taking many images among recurrent ranges.
REFERENCES
[1]

Clelland, W.D., Fens, T.W.: Automated rock

characterization with SEM/image-analysis
techniques. SPE Formation Evaluation, 6 (4),
437-443, 1991.

[2]

Tovey, N.K., Krinsley, D.H.: Mineralogical

mapping of scaning micrograps. Sedmentary
Geology, vol .75, pp 109-123, 1991.

[3]

Stirling, J. A. R.: Image analysis techniques

applied to potash ores. W. Petruk, (Ed.), In
Short Course on Image Analysis Applied to
Mineral and Earth Sciences pp 141-150, 1989.

Maximum Value
1
0.97
0.99
12.31
1

[4]

Bodziony, J., Konstankiewicz, K., Mlynarczuk,

M., Ratajczak, T.: Evaluation of the pore
structure in soil speciments, Acta Stereol,, 12,
pp 243-248, 1993.

[5]

Tovey, N. K., Hunslow, M. W.: Quantitative

micro-porosity and orientation analysis in soils
and sediments. Journal of the Geological
Society London, vol 152 no:1,
pp 119-129,
1995.

[6]

Meng, B.: Determination and interpretation of

fractal properties of the sandstone pore system.
Materials and Structures, 89, 188, pp 195-205,
1996.

[7]

Zheng, Z.: Compressive stress-induced

microcracks in rocks and applications to
seismic anisotropy and borehole stability. PHD
Thesis, Department of Materials Science and
Material Engineering Univesity of California,
Berkley, p 147, 1989

[8]

Goodchild, J. S., Fueten, F.: Edge detection in

petrographic images using the rotating polarizer
stage. Computer Geoscience, 24, 8, pp. 745751, 1998.

[9]

Jenkins, B. M., Boey, C. Y., Philips, P. L.

Appliying image analysis to the automatic
charactersation of dead-burnt magnesia. Proc.
Congr. on Applied Mineralogy, pp 26-52,
Pretoria, 1991.

[10] Friedman, G. M.: Comparison of moment

measures for siening andthin section data in
sedimentary petrological studies. J. Sediment
Petrol, 32, 1, 15-25, 1992.
[11] Persson, A. L.: Image analysis of fine aggregates
characterisation of shape grain-size parameters.,
Department of Civil and Enviromental
Engineering, Royal Institute of Technology,
Licentiate Thesis, Stockholm, p. 185, 1996.
[12] Wang, W.: Computer vision for aggregates.,
Department of Civil and Enviromental
Engineering, Royal Institute of Technology,
Doctoral Thesis, Stockholm, p. 104, 1997.
[13] Maerz, N. H.: Aggregate sizing and shape
determination using digital image processing.,
Center for Aggregates Research (ICAR) Sixth
Annual Symposium Proceedings, St Louis,
Missouri, April
pp 195-203, 1998.

8
[14] Kemeny, J., Kauda, R., Mofya, E., Lever, P.:
Improvements in blast fragmentation models
using digital image processing, Fragblast
Journal, 6, pp 311-320, 2002.
[15] L. Horovistiz, J. R. Frade and L. R. O. Hein. :
Comparison of Fracture Surface and Plane
Section Analysis for Ceramic Particle Size
Characterization. Journal of the European
Ceramic Society, 24, 4 , pp. 619-626.2004.
[16] C. Boukouvalas, J Kittler, R Marik, M
Mirmehdi and M Petrou.: Ceramic Tile
Inspection for Color and Structural defects.
I.E.E.E. Transactions on Pattern Analysis and
Machine Intelligence, vol. 14, no. 1 1998.
[17] Wang, W.: Image Analysis of Size And Shape
of Mineral Particles. Fourth International
Conference on Fuzzy Systems and Knowledge
Discovery, pp. 7-10, 2007
[18] Karakus, D.: Determination of the Structural
Properties of rocks with Image Analysing
Methods. The Graduate School of Natural and
Applied Sciences, PhD Thesis, zmir, p.173,
2006
[19] Gonzalez, C. R. ve Woods, E. R.: Digital image
processing, USA: Addison Wesley Publishing,
p. 976, 1993
[20] Reedy, C., L.,: Review of Digital Image Analysis
Of Petrographic Thin Sections in Conservation
Research. Journal of the American Institute for
Conservation, Vol. 45, No. 2 , pp. 127-146,
2006.
[21] Karakus, D., Onur, A. H., : Investgaton of
Determination of Rock Mass Properties by
Image Processing Methods. Geosound Science
and Technology Bulletin on Earth Science, vol.
48-49, pp. 237-255, 2006.
[22] Nordtest Method Standart.: Agregates: Size
Distribution, NT Build 486, pp. 3. Appoved
date: 1998.

Volume 12 - Issue 23