International Journal of Image Processing (IJIP)

Text-only Preview


VOLUME 6, ISSUE 1, 2012


ISSN (Online): 1985-2304
International Journal of Image Processing (IJIP) is published both in traditional paper form and in
Internet. This journal is published at the website, maintained by
Computer Science Journals (CSC Journals), Malaysia.

IJIP Journal is a part of CSC Publishers
Computer Science Journals


Book: Volume 6, Issue 1, February 2012
Publishing Date: 21-02- 2012
ISSN (Online): 1985-2304

This work is subjected to copyright. All rights are reserved whether the whole or
part of the material is concerned, specifically the rights of translation, reprinting,
re-use of illusions, recitation, broadcasting, reproduction on microfilms or in any
other way, and storage in data banks. Duplication of this publication of parts
thereof is permitted only under the provision of the copyright law 1965, in its
current version, and permission of use must always be obtained from CSC

IJIP Journal is a part of CSC Publishers

(c) IJIP Journal
Published in Malaysia

Typesetting: Camera-ready by author, data conversation by CSC Publishing Services - CSC Journals,

CSC Publishers, 2012


The International Journal of Image Processing (IJIP) is an effective medium for interchange of
high quality theoretical and applied research in the Image Processing domain from theoretical
research to application development. This is the forth issue of volume four of IJIP. The Journal is
published bi-monthly, with papers being peer reviewed to high international standards. IJIP
emphasizes on efficient and effective image technologies, and provides a central for a deeper
understanding in the discipline by encouraging the quantitative comparison and performance
evaluation of the emerging components of image processing. IJIP comprehensively cover the
system, processing and application aspects of image processing. Some of the important topics
are architecture of imaging and vision systems, chemical and spectral sensitization, coding and
transmission, generation and display, image processing: coding analysis and recognition,
photopolymers, visual inspection etc.

The initial efforts helped to shape the editorial policy and to sharpen the focus of the journal.
Starting with volume 6, 2012, IJIP appears in more focused issues. Besides normal publications,
IJIP intend to organized special issues on more focused topics. Each special issue will have a
designated editor (editors) - either member of the editorial board or another recognized specialist
in the respective field.

IJIP give an opportunity to scientists, researchers, engineers and vendors from different
disciplines of image processing to share the ideas, identify problems, investigate relevant issues,
share common interests, explore new approaches, and initiate possible collaborative research
and system development. This journal is helpful for the researchers and R&D engineers,
scientists all those persons who are involve in image processing in any shape.

Highly professional scholars give their efforts, valuable time, expertise and motivation to IJIP as
Editorial board members. All submissions are evaluated by the International Editorial Board. The
International Editorial Board ensures that significant developments in image processing from
around the world are reflected in the IJIP publications.

IJIP editors understand that how much it is important for authors and researchers to have their
work published with a minimum delay after submission of their papers. They also strongly believe
that the direct communication between the editors and authors are important for the welfare,
quality and wellbeing of the Journal and its readers. Therefore, all activities from paper
submission to paper publication are controlled through electronic systems that include electronic
submission, editorial panel and review system that ensures rapid decision with least delays in the
publication processes.

To build its international reputation, we are disseminating the publication information through
Google Books, Google Scholar, Directory of Open Access Journals (DOAJ), Open J Gate,
ScientificCommons, Docstoc and many more. Our International Editors are working on
establishing ISI listing and a good impact factor for IJIP. We would like to remind you that the
success of our journal depends directly on the number of quality articles submitted for review.
Accordingly, we would like to request your participation by submitting quality manuscripts for
review and encouraging your colleagues to submit quality manuscripts for review. One of the
great benefits we can provide to our prospective authors is the mentoring nature of our review
process. IJIP provides authors with high quality, helpful reviews that are shaped to assist authors
in improving their manuscripts.

Editorial Board Members
International Journal of Image Processing (IJIP)



Professor Hu, Yu-Chen
Providence University (Taiwan)


Professor. Khan M. Iftekharuddin
University of Memphis
United States of America

Assistant Professor M. Emre Celebi
Louisiana State University in Shreveport
United States of America

Assistant Professor Yufang Tracy Bao
Fayetteville State University
United States of America

Professor. Ryszard S. Choras
University of Technology & Life Sciences

Dr. Huiyu Zhou
Queen's University Belfast
United Kindom

Professor Yen-Wei Chen
Ritsumeikan University

Associate Professor Tao Gao
Tianjin University


Dr. C. Saravanan
National Institute of Technology, Durgapur West Benga

Dr. Ghassan Adnan Hamid Al-Kindi
Sohar University

Dr. Cho Siu Yeung David
Nanyang Technological University

Dr. E. Sreenivasa Reddy
Vasireddy Venkatadri Institute of Technology

Dr. Khalid Mohamed Hosny
Zagazig University

Dr. Chin-Feng Lee
Chaoyang University of Technology

Professor Santhosh.P.Mathew
Mahatma Gandhi University

Dr Hong (Vicky) Zhao
Univ. of Alberta

Professor Yongping Zhang
Ningbo University of Technology

Assistant Professor Humaira Nisar
University Tunku Abdul Rahman

Dr M.Munir Ahamed Rabbani
Qassim University

Dr Yanhui Guo
University of Michigan
United States of America

Associate Professor Andras Hajdu
University of Debrecen

Dr M.Munir Ahamed Rabbani
Qassim University

Assistant Professor Ahmed Ayoub
Shaqra University

Dr Irwan Prasetya Gunawan
Bakrie University

Assistant Professor Concetto Spampinato
University of Catania


Volume 6, Issue 1, February 2012

1 - 12
Improving Morphology Operation for 2D Hole Filling Algorithm
Mokhtar M. Hasan, Pramod K. Mishra

13 - 25
Unified Approach With Neural Network for Authentication, Security and Compression of
Dattatherya, S. Venkata Chalam, Manoj Kumar Singh

26 - 37
Face Recognition Using Improved FFT Based Radon by PSO and PCA Techniques
Hamid M. Hasan, Waleed A. AL.Jouhar, Majed A. Alwan

38 - 53
Computer Aided Visual Inspection of Aircraft Surfaces

Rafia Mumtaz, Mustafa Mumtaz, Atif Bin Mansoor, Hassan Masood

54 - 67
Data-Driven Motion Estimation With Spatial Adaptation

Alessandra Martins Coelho, Vania Vieira Estrela

International Journal of Image Processing (IJIP), Volume (6) : Issue (1) : 2012

Mokhtar M. Hasan & Pramod K. Mishra
Improving Morphology Operation for 2D Hole Filling Algorithm

Mokhtar M. Hasan

[email protected]
Computer Science Department/ Faculty of Science
Banaras Hindu University
Varanasi, 221005, India

Pramod K. Mishra

[email protected]
Computer Science Department/ Faculty of Science
Banaras Hindu University
Varanasi, 221005, India


Object detection may result with some noises, the correct detecting of such object plays a major
role for later recognition steps, the interior noise of the object must be removed, the
morphological operations are used successfully for this purpose, these morphological operations
are applied for 2D holes filling using dilation operation, in this paper we have enhanced this
algorithm to get better and faster version that will reduce the processing time dramatically by
using a dynamic marker, we have applied two kind of markers with different structuring elements
but same size which is 3x3 and those markers are used according to the structure of the sub-
window of the object, the processing time is reduced, and our algorithm reduces this time
approximately to one third, the results also enhanced since there are some cases missed by the
extant version of morphological operations holes filling algorithm.

Keywords: Morphological Operation, Dilation Morphological, Object Filling, 2D Object Filling,
Object Holes, Noise Removal, Scanning 2D Objects.

We must start with the formal definition of the hole, cavity, and concavity, the hole has been
defined from different views but all referring to the same common ground, we can define the hole
as "a background region that is a subset of a object", since the hole region should be a subset of
their object that belongs to, the Gonzalez and Woods definition in [1] was verbatim "background
region surrounded by a connected border of foreground pixels"
, the other definition that found in
Oxford Dictionary [2] was "a hollow space in something solid or in the surface of something", in
all of the mentioned definitions; we have to exclude the cup handle from being a hole since it is
part of the cup [3] or the area that lies in the middle of a doughnut [3] as well, the 3D definition of
the hole can be taken from [3] which is "break in a surface mesh".

The cavity and concavity are dedicated for 3D objects since these terms has a more
understandable meaning, the Oxford Dictionary [2] for this term verbatim is "a hole or empty
space inside something solid
", so, we have moved to the space view of the hole, the other
definition [4] is the cavity is the existence of a hollow [2] in an object, or verbatim from [4] "a
bounded connected component of the background
", while the concavity is considered to be a
contour with concave shape [2].

Object scanning may produce some missing portions or missing regions even pixels of the
scanned object due to and hence, we address the hole filling algorithms to close up and fill out
this missing information.

There are many usages of the holes filling algorithms, the preprocessing steps may add some
extra holes [5] to the processed object since the segmentation operation is not always perfect, so,
International Journal of Image Processing (IJIP), Volume (6) : Issue (1) : 2012

Mokhtar M. Hasan & Pramod K. Mishra
applying the hole filling will eliminate the original holes and the new added holes as well [5]. Other
usage is for Cultural Heritage [6] which has several reasons, the authors in [6] summarize those
reasons by the sampling accuracy [6], speed gained from scanning technique [6], and acquiring
these object data without touching the object [6] since some objects are very fragile due to their

Other factors that impact upon the producing of the holes are occlusion [5], low reflectance [5] or
the original scanned object has some missing portions [5].

Furthermore, in order to produce a best model that fits the original object [7] we need to remove
any noises that classified as object region or foreground region.

In this paper we have modify the morphology operation for object filling and we gain a more faster
and more enhanced results since there are some missing cases by the original algorithm. This
paper is organized as follows: section 2 lists some related work associated before, part 3 gives a
review for the meaning of the morphological operations, part 4 shows the other algorithms applied
in this field as well as the original version of the algorithm that we have modified, part 5 shows our
algorithm, part 6 explains the impact of the selection of the structuring elements that we have
made to overcome some shortcoming issues that arose from the original version of the extant
algorithm, part 7 shows the experimental results, part 8 shows the speed factor comparisons
between the old version and our version, and finally, part 9 gives the conclution.

Gonzalez and Woods [1] presented two different algorithms for object hole filling, these algorithm
use the morphological dilation operation for such purpose, the first algorithm is restricted since an
initial pixel within the object hole must be specified, which is more difficult in real time automatic
applications that needs the filling to be done completely without human interference.

The other algorithm presented by them was applied by dilation algorithm as well but this time no
initial pixel needed to be located. They started with an initial marker which is the complement of
the border of the original input image and black pixels elsewhere, and their 3x3 structuring
element (SE for shorthand) was all ones, and by applying the algorithm, the border comes and
closing the objects as every epoch of the algorithm until there is no more changes in the resulted
marker, and by anding logical operation between the complement of the marker and the original
image matrix, they got the image holes matrix and they got the filled image by oring logical
operation between the last two matrices.

The other approach was applied by us in [7] which depends on finding a connected region around
the background pixel to be recognized as hole pixel, any initials or prior assumption have not
required, they convolved a diamond filter with every background pixel and a roadmap is created,
then they trace this road map to find a connected path which means the background pixel has a
connected region surrounds it, if such path existed; then this hole pixel as foreground pixel is
considered, otherwise, background pixel is considered .

The term of morphology came from the biology branch [1] which means the study of the structure
of the animals and plants [1], the same term can applied to image objects for studying the
structure of these objects in the image [1], there are two basic and important operations which are
the dilation and erosion [1], the Oxford Definition [2] for dilation is "to become or to make
something larger, wider or more open
", and the Oxford Definition [2] for erosion is "to gradually
destroy the surface of something through the action of wind, rain, etc
", and in the image erosion ,
the object is eroded by the action of filters which is called the SE that has a reference point that
dominate the matching with the object window since we have no wind nor rain here.

International Journal of Image Processing (IJIP), Volume (6) : Issue (1) : 2012

Mokhtar M. Hasan & Pramod K. Mishra
3.1 Morphology Operation Notations
Let A Z2, which is the 2D space of the (x, y), and also let B Z2 be the SE which controls the
structure of the morphological operations, then for binary images, dilation and erosion can be
defined as (1) and (2) respectively:

A B = { z | z = a . b , a A & b B } (1)

A B = { z | z = a . b , a A & b B } (2)

Where and represent the dilation and erosion operations respectively, and . (dot)
represents the logical anding operation, Equation (1) says that the output of the dilation operation
is set to reference point if there is any hit between object A and SE B, this reference point
represents the output value if such hit existed, and for Equation (2); all SE must be included
inside the object A to produce the reference point as an output which means fit.

The other term is the reconstruction which may exploit the dilation or erosion to get its process
done; we will focus on dilation since it is used in this paper, the formulation for this operation as in
(3) which taken from [1]:


Where R denotes for reconstruction operation, D for dilation, A is the mask image, F is the
marker and k represents the number of repeated times until (4) holds:


We will define no more operations since it is out of the scope of this paper, any more definition
can be referred to [1,8].

The formulation of the two morphology filling operation can be formulated as follow:

4.1 Hole-Pixel Initial Algorithm (HPIA)
Consider O = { o0, o1, .., ok | oi represents the (x, y) pixels of object i, k represents the number of
objects in image }, then the initial for this algorithm is a vector P where in (5) and (6) :

P={Ii | i=1, 2, ... k}


Ii = {(xj, yj) , j number of holes in object i } (6)

This required having starting points that belongs for each hole in each object of the image, which
is very difficult since the real time application for such algorithm cannot go along with this since it
requires the human interaction in a very essential step which is all the latter processing depends
on. However, the The equation for this algorithm is in (7) taken from [1] which closes the holes of
the object after certain number of epochs which means until there is no change in Xk:

X k = (X k-1 B ) Ac , for k = 1, 2, 3, ... (7)

Where X0 is the marker with initial points Ii , B is the 3x3 SE with zeros at the corners and ones
elsewhere, and A is the original input binary image.
International Journal of Image Processing (IJIP), Volume (6) : Issue (1) : 2012