International Journal of Computer Science, Engineering and Information Technology (IJCSEIT)

International Journal of Computer Science, Engineering and Information Technology (IJCSEIT), Vol.3,No.4,August 2013
DOI : 10.5121/ijcseit.2013.3402 11
OBJECT SEGMENTATION USING MULTISCALE
MORPHOLOGICAL OPERATIONS
Dr. A. Srikrishna1
, P. Pallavi2
, V. Geetha Madhuri3
, N. Neelima4
1,2,3,4
Department of Information Technology,
RVR & JC College of Engineering,
Chowdavaram, Guntur, Andhra Pradesh, India
atlurisrikrishna@yahoo.com
paladugupallavi3@gmail.com
madhurigeetha33@gmail.com
neelima_nalla@yahoo.co.in
ABSTRACT
Object segmentation plays an important role in human visual perception, medical image processing and
content based image retrieval. It provides information for recognition and interpretation. This paper uses
mathematical morphology for image segmentation. Object segmentation is difficult because one usually
does not know a priori what type of object exists in an image, how many different shapes are there and
what regions the image has. To carryout discrimination and segmentation several innovative segmentation
methods, based on morphology are proposed. The present study proposes segmentation method based on
multiscale morphological reconstructions. Various sizes of structuring elements have been used to segment
simple and complex shapes. It enhances local boundaries that may lead to improve segmentation accuracy.
The method is tested on various datasets and results shows that it can be used for both interactive and
automatic segmentation.
KEYWORDS
Morphology, Structuring Element, Segmentation, Edge Detection, Skeletanization
1. INTRODUCTION
Humans recognize various objects in an image though the objects may vary somewhat in different
viewpoints and on various transformations. Object segmentation is useful task in object
recognition. The object recognition determines an object in a given set of objects in an image or
image sequence. In order to perform object recognition the objects from a give image or image
sequence are to be identified. For this object segmentation that is to distinguish objects from
background is performed.
Object segmentation [3] is the process of partitioning a digital image into multiple segments (sets
of pixels, also known as super pixels). The goal of segmentation is to simplify and/or change the
representation of an image into something that is more meaningful and easier to analyze. Object
segmentation is typically used to locate objects and boundaries (lines, curves, etc.) in images.
More precisely, object segmentation is the process of assigning a label to every pixel in an image
such that pixels with the same label share certain visual characteristics.
Some of the applications of object segmentation are content based image retrieval, machine
vision, medical imaging, object detection, recognition tasks, traffic control systems, video

12
surveillance. Object segmentation is the partition of an image into a set of non overlapping
regions whose union is the entire image. The purpose of segmentation is to decompose the image
into part that is meaningful with respect to a particular application. Concerning visual signal
processing, image segmentation is essential for various applications. It describes the process
whereby each pixel in an image is labeled, such that pixels with the same label present coherent
visual characteristics. This allow for a semantic approach to image analysis. One way to perform
image segmentation is to simply utilize the clustering algorithm in the color space domain [1],
i.e., HSV or RGB; segmentation can also be based on the statistics of the color space description
of the image, e.g., color histogram. These methods are carried out in the color space domain
instead of the image pixel domain, whose results depend on the initial cluster setting. Edge-based
segmentation is simple but it requires a further linking procedure to segment an image [2], [3],
[4]. Among color region-based approaches, the region-growing approach [5] provides an initial
set of seeds; regions are then grown by comparing neighboring pixels, which are merged [6] with
the region with the closest mean color. JSEG [7], [8], [9], [10] seeks to divide an image into
spatially continuous disjoint and homogenous regions based on the image. It uses a region
merging approach, but the color information between entire neighboring regions, rather than
individual pixels, is utilized. Experiments show that JSEG provides satisfactory results on most
color images. The watershed technique splits one image into regions based on its gray-level
topology and is performed on the gradient image. Regions are split by watersheds, which are
constructed from adjacent catchment basins. Although it has the advantage of being able to
segment regions with closed contours, it suffers from over segmentation and requires region
merging processing afterward. Multiscale morphological reconstruction [11] is used to eliminate
the over segmentation in the watershed algorithm. Graph based segmentation [12], [13] takes the
global image information and local spatial relationships into consideration to perform image
segmentation. It defines a predicate to measure the boundary evidence between two neighboring
regions to yield a graph-based representation of one image.
The general-purpose segmentation algorithms [1], [7], [8], [9], [10], [12], [13] represent one
image with disjoint regions of homogeneous color/texture features for higher level applications,
and the object segmentation ones [11], [14] extract image objects with different gray level
variations and noise attacks. The former does not address object segmentation from its design
target. The latter focuses on segmenting the object of different scale but do not perform parameter
adaptation when dealing with images with different background (BG) variation and object
contents. To utilize both region and object-based segmentation capabilities to handle the object
segmentation for large-scale database images in a robust and principled manner an algorithm is
proposed. The proposed algorithm is known as object segmentation using multiscale morphology
(OSMM). Morphological open (close) by reconstruction [14], [15], i.e., OR (CR), with a proper
structure element (SE) on the gray levels, to automatically segment images’ object region is used.
Section 2 discusses the basic terminology of mathematical morphology. Section 3 discusses
segmentation using multiscale morphology. Section 4 deals with results and discussions.
2. BASIC TERMINOLOGY OF MATHEMATICAL MORPHOLOGY
2.1. Mathematical Morphological Operations
2.1.1. Dilation
Dilation is one of the elementary operators of mathematical morphology, that is, it is a building
block for a large class of operators. The key process in the dilation operator is the local
comparison of a shape, called structural element, with the object to be transformed. When the
structural element is positioned at a given point and it touches the object, then this point will
appear in the result of the transformation, otherwise it will not. In dilation the value of the output

13
pixel is the maximum value of all the pixels in the input pixel's neighbourhood. In a binary image,
if any of the pixels is set to the value 1, the output pixel is set to 1. The dilation of a gray level
image I(x, y) by two-dimensional structuring element B is defined as follows
(1)
2.1.2. Erosion
Erosion is one of two fundamental operations (the other being dilation) in morphological image
processing from which all other morphological operations are defined. It was originally defined
for binary images, later being extended to gray scale images, and subsequently to complete
lattices.
The erosion is one of the elementary operators of mathematical morphology, that is, it is one of
the building blocks of a large class of operators. The key mechanism under the erosion operator is
the local comparison of a shape, called structural element, with the object that will be
transformed. If, when positioned at a given point, the structural element is included in the object
then this point will appear in the result of the transformation, otherwise not. The value of the
output pixel is the minimum value of all the pixels in the input pixel's neighbourhood. The
erosion of a gray level image I(x, y) by two dimensional structuring element B is defined as
follows
(2)
2.1.3. Opening
Opening smoothes the contour of an object, breaks narrow isthmuses, and eliminates thin
protrusions. The opening of I by structuring element B is obtained by the erosion of I by B,
followed by dilation of the resulting image by B is denoted by I○B which is given in the form of
equation as follows
(3)
2.1.4. Closing
Closing tends to smooth sections of contours but, as opposed to opening, it generally fuses narrow
breaks and long thin gulfs, eliminates small holes, and fills gaps in the contour. The closing of I
by B is obtained by the dilation of I by B, followed by erosion of the resulting structure by B is
denoted by I●B which is given in the form of equation as follows
(4)
2.2. Multiscale Morphological Operations
Let the structuring element defined in Equations 1, 2, 3, 4 be where denotes the smallest
structuring element size in the discrete domain. The homothetic of a convex structuring
element can be obtained by dilating recursively times with itself given by Equation 5.
(5)

14
By controlling , the multiscale morphological operations decompose one image into a set of
filtered images. These operations are self-calibrated in which the filtered image produced by a
structuring element of a particular scale should strictly contain only the features of that scale.
Multiscale morphological reconstruction operations on gray-level images can be applied in a dual
approach to segment objects. Let , and the morphological open by reconstruction (OR)
can be defined by Equation 6.
(6)
Where
Similarly, let the morphological close by reconstruction (CR) can be defined by
Equation 7.
(7)
where
3. SEGMENTATION USING MULTISCALE MORPHOLOGY
The general-purpose segmentation algorithms such as mean shift segmentation, edge based
segmentation represent one image with disjoint regions of homogeneous color or texture features
for higher level applications, and the object segmentation ones such as watershed segmentation,
graph based segmentation extract image objects with different gray level variations and noise
attacks. The former does not address object segmentation from its design target. The latter
focuses on segmenting the object of different scale but do not perform parameter adaptation when
dealing with images with different background (BG) variation and object contents. In order to
overcome the drawbacks, segmentation using multiscale morphology has been proposed. This
method utilizes both region- and object-based segmentation capabilities to handle the object
segmentation in a robust and principled manner. Figure 1 shows the block diagram of the object
segmentation using multiscale morphological reconstructions.
Figure 1. Object Segmentation using Multiscale Morphology.

15
3.1. Background Variations
Open by reconstruction (OR),close by reconstruction (CR) and top- (bottom-) hat operations are
used to segment object regions and their gray levels , where ⋀ is a binary AND
operation. The reconstruction operations of OR (CR) would not be fully iterated for stable
outcome so that convex (concave) gray-level variation regions can be located. The image
processed by this partial OR (CR) operation is denoted as with to be
distinguished from the fully reconstructed one in that requires iterations. The basic idea is
images with identifiable back ground regions (BGs) usually present homogeneously evolving
gray-level BGs, i.e. , is continuous. When performing gray-level open by
reconstruction (OR) operations on image with suitable structuring element sizes most background
(BG) regions would coincide with the processed image . To precisely locate back ground
regions (BGs) the structuring element size in should be properly selected such that object
boundaries are identifiable. For precisely locating the structuring element, structuring element
size is gradually enlarged and the frame difference of gray levels between I and i.e.
which is given by Equation 8 is calculated ( difference of gray levels for consecutive
structuring elements is calculated).
(8)
The threshold value is calculated by using the Equation 9
(9)
where is the size of the image.
To find a proper structuring element such that it yields a nearly stable gray-level variation, i.e.,
threshold given by Equation 9 should nearly approach zero. The proper has threshold given by
Equation 9 values between 0 and 1, and different images would require different to
achieve the aforementioned stability due to differing object contents.
3.2. Edge Detection
The edge pixels of the input image I are detected by using morphological operations. The edge
pixels denoted by Edge(I) for an image I is given by the Equation 10
(10)
where Dilation(I), Erosion (I) are dilation, erosion operations on input image I as defined in
Equations 1, 2 respectively.
3.3. Skeletanization
The skeleton of an object L(I) can be obtained by edge image and background variations. The
skeleton obtained using skeletanization process is given by Equation 11.
(11)

16
3.4. Segmentation
The skeleton obtained using skeletanization process is given as the input for reconstruction of
object. In the reconstruction the input image is dilated by varying sizes of structuring element B
which satisfies the condition using threshold value given by Equation 9 which is in the range of 0
and 1. Let be the outputs of dilation for varying sizes of structuring elements.
Then the segmented object is given by Equation 12.
(12)
The overall process of segmentation is given in the Algorithm
3.5. Algorithm Object Segmentation using Multiscale Morphology (OSMM)
Step 1: Read the input image.
Step 2: Convert the image into gray scale image.
Step 3: Perform “opening(erosion followed by dilation) or closing (dilation followed by
erosion)” of the input image using varying sizes of structuring elements.
Step 4: The structuring element size is gradually enlarged and the frame difference of gray levels
between I and i.e. which is given by Equation 8 is recorded.
Step 5: Calculate the threshold by using the formula in Equation 9.
Step 6: The structuring elements whose threshold values are between 0 and 1 are stored.
Step 7: Calculate the edge pixels of the input image by using the formula in Equation 10.
Step 8: Perform detaching process given by Equation 11.
Step 9: Perform reconstruction of segmented by using Equation 12.
4. RESULTS AND DISCUSSIONS
We evaluate shape prototypes in the context of object segmentation. These object segmentation
techniques are used as a pre processing step for object recognition. We begin with a set of 200
objects, representing distinct views of collection of 101 objects from Caltech101 database. We are
using this step as a pre processing step for object recognition so the input image consists of only
one object.
To evaluate the performance of the algorithm correctness and completeness criteria are
considered which are defined by Equations 13 and 14 respectively.
Correctness can be defined as the percentage of correctly extracted region (ground truth) by the
segmentation algorithm and can be calculated using Equation 13.
(13)
Completeness can be defined as the percentage of the ground truth region extracted by the
segmentation algorithm and can be calculated using Equation 14.
(14)

17
where to obtain the true positive (TP) image a logical AND operation was performed between the
ground truth and the resultant image. The difference between the ground truth image and the true
positive image was taken as the false negative (FN) image of the respective segmented image.
The difference between the segmented image and the true positive image was taken as the false
positive (FP) image of the respective segmented image. The TP, FP, FN are illustrated in Figure
2. Here we consider the ground truth which is a manually segmented image. The ground truth
images are compared with the image segmented with the proposed algorithm. The Figure 3 to
Figure 8 shows that the segmented images using proposed algorithm are very similar to ground
truth. The objective analysis of the segment is evaluated using correctness and completeness
which is shown from Table 1 to Table 6. The correctness is above 80% in most of the cases and
completeness is above 90% in all cases.
Figure 2. Ground truth region of object and segmented region
Lilly.png
Figure 3. a. Shows the Lilly image. b. Shows the ground truth which is taken manually. c-f represent the
output of the image with varying structuring elements.

18
Table 1. Correctness and Completeness values for various sizes of structuring elements for Lilly.png.
Scissors.jpg
Figure 4. a. Shows the Scissors image. b. Shows the ground truth which is taken manually. c-f represent the
output of the image with varying structuring elements
Table 2. Correctness and Completeness values for various sizes of structuring elements for Scissors.jpg
Structuring Element
Size
Correctness Completeness
5x5 86.8629 90.5704
7x7 81.7156 93.7795
9x9 77.2008 95.2278
11x11 75.0547 96.3634
13x13 72.6632 97.4097
17x17 67.6016 98.7240
Structuring
Element
Size
Correctnes
s
Completenes
s
7x7 86.3627 79.4165
9x9 82.1526 85.8422
11x11 79.8946 88.6663
13x13 78.5122 91.3224
15x15 78.129 95.0174
17x17 74.5465 97.4122

19
Mandolin.jpg
Figure 5. a. Shows the Mandolin image. b. Shows the ground truth which is taken manually. c-f represent
the output of the image with varying structuring elements.
Table 3. Correctness and Completeness values for various sizes of structuring elements for Mandolin.jpg
Structuring
Element
Size
Correctnes
s
Completenes
s
5x5 91.1111 90.4906
7x7 86.8741 94.5327
9x9 83.3252 96.9289
11x11 81.4175 98.4424
13x13 78.5678 99.1486
15x15 75.0938 99.6657

20
Seahorse.jpg
Figure 6. a. Shows the Seahorse image. b. Shows the ground truth which is taken manually. c-f represent
Table 4. Correctness and Completeness values for various sizes of structuring elements for Seahorse.jpg
Structuring
Element
Size
Correctnes
s
Completenes
s
7x7 90.0999 93.7143
9x9 85.1340 96.0395
11x11 83.0496 96.8145
13x13 81.0013 97.2830
15x15 78.7993 97.5981
17x17 76.5526 98.1602

21
Flamingo.jpg
Figure 7. a. Shows the Flamingo image. b. Shows the ground truth which is taken manually. c-f represent
the output of the image with varying structuring elements
Table 5. Correctness and Completeness values for various sizes of structuring elements for Flamingo.jpg
Structuring
Element
Size
Correctnes
s
Completenes
s
5x5 85.456 85.3091
7x7 81.234 92.5348
9x9 77.6136 96.1200
11x11 74.9444 97.2496
13x13 73.3578 97.90656
15x15 71.6859 98.6985

22
Dragonfly.jpg
Figure 8. a. Shows the Dragonfly image. b. Shows the ground truth which is taken manually. c-f represent
Table 6. Correctness and Completeness values for various sizes of structuring elements for Dragonfly.jpg
5. CONCLUSIONS
A simple and regular image object segmentation method has been proposed to deal with large-
scale image databases. It performs dual multiscale morphological reconstruction operations on the
gray levels of entire images to identify the objects. Experiments have demonstrated that OSMM
yields better image object segmentation accuracy, both on shape region and boundary. The results
show that the correctness and completeness of the image increases as the structuring element
increases. The segmentation process can be used for identification of object in any database.
Structuring
Element
Size
Correctnes
s
Completenes
s
5x5 84.6530 97.1433
7x7 80.0787 99.3071
9x9 76.1462 99.6933
11x11 74.6419 99.7501
13x13 71.6884 99.8012
15x15 69.0193 99.8523

23
ACKNOWLEDGEMENTS
The authors would like to express their gratitude to Dr.K.Basavapunnaiah, President, Sri. R.
Gopala Krishna, Secretary & Correspondent, and Dr. M. Gopala Krishna, Treasurer, RVR & JC
College of Engineering for providing necessary research infrastructure. They would like to thank
Dr. A. Sudhakar, Prinicipal for his invaluable suggestions and constant encouragement which led
to improvise the presentation of this paper.
REFERENCES
[1] D. COMANICU AND P. MEER, “MEAN SHIFT A ROBUST APPROACH TOWARD FEATURE SPACE
ANALYSIS,” IEEE TRANS. PATTERN ANAL. MACH. INTELL., VOL. 24, NO. 5, PP. 603–619, MAY 2002.
[2] M. P. PATHEGAMA AND Ö. GÖL, “EDGE-END PIXEL EXTRACTION FOR EDGE BASED IMAGE
SEGMENTATION,” IN PROC. WORLD ACAD. SCI. ENG. TECHNOL., JAN. 2005, VOL. 2, PP. 164–167.
[3] RAFAEL C. GONZALEZ AND RICHARD E. WOODS, DIGITAL IMAGE PROCESSING, 2ND EDITION,
PEARSON EDUCATION, INC, INDIA, 2002.
[4] M. P. PATHEGAMA AND Ö. GÖL, “AN ARTIFICIAL NEURAL PROCESS TO CREATE CONTINUOUS OBJECT
BOUNDARIES IN MEDICAL IMAGE ANALYSIS”, INTERNATIONAL SCIENTIFIC JOURNAL OF COMPUTING,
VOL. 3, ISSUE 1, 2004.
[5] P. SALEMBIER AND F. MARQUES, “REGION-BASED REPRESENTATIONS OF IMAGE AND VIDEO-
SEGMENTATION TOOLS FOR MULTIMEDIA SERVICES,” IEEE TRANS.CIRCUITS SYST. VIDEO TECHNOL.,
VOL. 9, NO. 8, PP. 1147–1169, DEC. 1999.
[6] F. MOSCHENI, S. BHATTACHARJEE, AND M. KUNT, “SPATIO-TEMPORAL SEGMENTATION BASED ON
REGION MERGING,” IEEE TRANS. ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 20, NO.
9, P. 897-915,1998.
[7] Y. DENG AND B. S. MANJUNATH, “UNSUPERVISED SEGMENTATION OF COLOR-TEXTURE REGIONS IN
IMAGES AND VIDEO,” IEEE TRANS. PATTERN ANAL. MACH. INTELL., VOL. 23, NO. 8, PP. 800–810,
AUG. 2001.
[8] L. G. UGARRIZA, E. SABER, S. R. VANTARAM, V. AMUSO, M. SHAW, AND R.BHASKAR, “AUTOMATIC
IMAGE SEGMENTATION BY DYNAMIC REGION GROWTH AND MULTIRESOLUTION MERGING,” IEEE
TRANS. IMAGE PROCESS., VOL. 18, NO. 10, PP. 2275–2288, OCT. 2009.
[9] S. BELONGIE, C. CARSON, H. GREENSPAN, AND J. MALIK, “COLOR- AND TEXTURE-BASED IMAGE
SEGMENTATION USING EM AND ITS APPLICATION TO CONTENT-BASED IMAGE RETRIEVAL,” PROC. OF
INTL. CONF. ON COMPUTER VISION, PP. 675-82, 1998.
[10] D.K. PANJWANI AND G. HEALEY, “MARKOV RANDOM FIELD MODELS FOR UNSUPERVISED
SEGMENTATION OF TEXTURED COLOR IMAGES,” IEEE TRANS. ON PATTERN ANALYSIS AND MACHINE
INTELLIGENCE, VOL. 17,NO.10, PP. 939-54, 1995.
[11] K. NALLAPERUMAL, K. KRISHNAVENI, J. VARGHESE, S. SAUDIA, S. ANNAM,AND P. KUMAR, “A
NOVEL MULTI-SCALE MORPHOLOGICAL WATERSHED SEGMENTATION ALGORITHM,” INT. J. IMAG. SCI.
ENG., VOL. 1, NO.2, PP. 60–64, APR.2007.
[12] P. F. FELZENSZWALB AND D. P. HUTTENLOCHER, “EFFICIENT GRAPH-BASED IMAGE SEGMENTATION,”
INT. J. COMPUT. VIS., VOL. 59, NO. 2, PP. 167–181, SEP. 2004.
[13] J. SHI AND J. MALIK, “NORMALIZED CUTS AND IMAGE SEGMENTATION,” IEEE TRANS. ON PATTERN
ANALYSIS AND MACHINE INTELLIGENCE, VOL. 22, NO. 8, P. 888-905, 2000.
[14] S. MUKHOPADHYAY AND B. CHANDA, “MULTISCALE MORPHOLOGICAL SEGMENTATION OF GRAY-
SCALE IMAGES,” IEEE TRANS. IMAGE PROCESS., VOL. 12, NO.5, PP. 533–549, MAY 2003.
[15] J. SERRA AND L. VINCENT, “AN OVERVIEW OF MORPHOLOGICAL FILTERING,” CIRCUITS SYST. SIGNAL
PROCESS., VOL. 11, NO. 1, PP. 47–108, MAR. 1992.

International Journal of Computer Science, Engineering and Information Technology (IJCSEIT)

More Related Content

What's hot (16)

Similar to International Journal of Computer Science, Engineering and Information Technology (IJCSEIT) (20)

Recently uploaded (20)

International Journal of Computer Science, Engineering and Information Technology (IJCSEIT)