Advances in Multimedia - An International Journal (AMIJ) Volume: 4 Issue: 5

---

ADVANCES IN MULTIMEDIA - AN
INTERNATIONAL JOURNAL (AMIJ)







VOLUME 2, ISSUE 1, 2011

EDITED BY
DR. NABEEL TAHIR








ISSN (Online): 2180-1223
Advances in Multimedia - An International Journal is published both in traditional paper form and
in Internet. This journal is published at the website http://www.cscjournals.org, maintained by
Computer Science Journals (CSC Journals), Malaysia.


AMIJ Journal is a part of CSC Publishers
Computer Science Journals
http://www.cscjournals.org

---

ADVANCES IN MULTIMEDIA - AN INTERNATIONAL JOURNAL
(AMIJ)

Book: Volume 2, Issue 1, March 2011
Publishing Date: 04-04-2011
ISSN (Online): 2180-1223

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
Publishers.



AMIJ Journal is a part of CSC Publishers
http://www.cscjournals.org

© AMIJ Journal
Published in Malaysia

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



CSC Publishers, 2011

---

EDITORIAL PREFACE

This is first issue of volume two of the Advances in Multimedia - An International Journal (AMIJ).
AMIJ is an International refereed journal for publication of current research in computer science
and computer security technologies. AMIJ publishes research papers dealing primarily with the
technological aspects of computer science in general and computer security in particular.
Publications of AMIJ are beneficial for researchers, academics, scholars, advanced students,
practitioners, and those seeking an update on current experience, state of the art research
theories and future prospects in relation to computer science in general but specific to computer
security studies. Some important topics cover by AMIJ are Animation, Computer Vision,
Multimedia Signal Processing, Visualization, Scanning, Multimedia Analysis, Multimedia
Retrieval, Motion Capture and Synthesis, Displaying, Dynamic Modeling and Non-Photorealistic
Rendering, etc.

The initial efforts helped to shape the editorial policy and to sharpen the focus of the journal.
Starting with volume 5, 2011, AMIJ appears in more focused issues. Besides normal publications,
AMIJ 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.

This journal publishes new dissertations and state of the art research to target its readership that
not only includes researchers, industrialists and scientist but also advanced students and
practitioners. The aim of AMIJ is to publish research which is not only technically proficient, but
contains innovation or information for our international readers. In order to position AMIJ as one
of the top ADVANCES IN MULTIMEDIA - AN INTERNATIONAL JOURNAL, a group of highly
valuable and senior International scholars are serving its Editorial Board who ensures that each
issue must publish qualitative research articles from International research communities relevant
to Advance Multimedia fields.

AMIJ 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 AMIJ. 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. AMIJ provides authors with high quality, helpful reviews that are shaped to assist
authors in improving their manuscripts.


Editorial Board Members
Advances in Multimedia - An International Journal (AMIJ)

---

TABLE OF CONTENTS




Volume 2, Issue 1, March 2011



Pages
1 - 17
Extended Performance Appraise of Image Retrieval Using the Feature Vector as Row
Mean of Transformed Column Image
Dr. H. B. Kekre, Sudeep D. Thepade & Akshay Maloo

18 - 29
Adaptive Sliding Piece Selection Window for BitTorrent Systems
Ahmed B. Zaky, May A. Salama & Hala H .Zayed













Advances in Multimedia - An International Journal (AMIJ), Volume (2), Issue (1)

---

Dr. H. B. Kekre, Sudeep D. Thepade & Akshay Maloo
Extended Performance Appraise of Image Retrieval Using the
Feature Vector as Row Mean of Transformed Column Image


Dr. H. B. Kekre




hbkekre@yahoo.com
Senior Professor, Computer Engineering,
MPSTME, SVKM’S NMIMS University,
Mumbai, 400056, India

Sudeep D. Thepade




sudeepthepade@gmail.com
Ph.D. Research Scholar and Associate Professor,
Computer Engineering,
MPSTME, SVKM’S NMIMS University,
Mumbai, 400056, India

Akshay Maloo




akshaymaloo@gmail.com
Student, Computer Engineering,
MPSTME, SVKM’S NMIMS University
Mumbai, 400056, India

Abstract

The extension to the content based image retrieval (CBIR) technique based on row mean of
transformed columns of image is presented here. As compared to earlier contemplation three
image transforms, now the performance appraise of proposed CBIR technique is done using
seven different image transforms like Discrete Cosine Transform (DCT), Discrete Sine Transform
(DST), Hartley Transform, Haar Transform, Kekre Transform, Walsh Transform and Slant
Transform. The generic image database with 1000 images spread across 11 categories is used
to test the performance of proposed CBIR techniques. For each transform 55 queries (5 per
category) were fired on the image database. Every technique is tested on both the color and grey
version of image database. To compare the performance of image retrieval technique across
transforms average precision and recall are computed of all queries. The results have shown the
performance improvement (higher precision and recall values) with proposed methods compared
to all pixel data of image at reduced computations resulting in faster retrieval in both gray as well
as color versions of image database. Even the variation of considering DC component of
transformed columns as part of feature vector and excluding it are also tested and it is found that
presence of DC component in feature vector improvises the results in image retrieval. The
ranking of transforms for performance in proposed gray CBIR techniques with DC component
consideration can be given as DST, Haar, Hartley, DCT, Walsh, Slant and Kekre. In color variants
of proposed techniques with DC component, the performance ranking of image transforms
starting from best can be listed as DCT, Haar, Walsh, Slant, DST, Hartley and Kekre transform.

Keywords- CBIR, DCT, DST, Haar, Walsh, Kekre, Slant, Hartley,Row Mean.


1. INTRODUCTION
The hefty sized image databases which are being generated from a variety of sources (digital
camera, video, scanner, the internet etc.) have posed technical challenges to computer systems to
store/transmit and index/manage image data effectively to make such large collections easily
accessible. Storage and transmission challenges are taken care by Image compression. The
challenges of image indexing are studied in the context of image database [2,6,7,10,11], which has
become one of the most important and promising research area for researchers from a wide range
of disciplines like computer vision, image processing and database areas. The need for faster and
better image retrieval techniques is increasing day by day. Some of important applications for
CBIR technology could be identified as art galleries [12,14], museums, archaeology [3],
Advances in Multimedia - An International Journal (AMIJ), Volume (2) : Issue (1) : 2011
1

---

Dr. H. B. Kekre, Sudeep D. Thepade & Akshay Maloo
architecture design [8,13], geographic information systems [5], trademark databases [21,23],
weather forecast [5,22], medical imaging [5,18], criminal investigations [24,25], image search on
the Internet [9,19,20].

1.1 Content Based Image Retrieval

In literature the term content based image retrieval (CBIR) has been used for the first time by Kato
et.al. [4], to describe his experiments into automatic retrieval of images from a database by color
and shape feature. The typical CBIR system performs two major tasks [16,17]. The first one is
feature extraction (FE), where a set of features, called feature vector, is generated to accurately
represent the content of each image in the database. The second task is similarity measurement
(SM), where a distance between the query image and each image in the database using their
feature vectors is used to retrieve the top “closest” images [16,17,26].
For feature extraction in CBIR there are mainly two approaches [5] feature extraction in spatial
domain and feature extraction in transform domain. The feature extraction in spatial domain
includes the CBIR techniques based on histograms [5], BTC [1,2,16], VQ [21,25,26]. The
transform domain methods are widely used in image compression, as they give high energy
compaction in transformed image [17,24]. So it is obvious to use images in transformed domain for
feature extraction in CBIR [23,28]. But taking transform of image is time consuming. Reducing the
size of feature vector by applying transform on columns of the image and finally taking row mean
of transformed columns and till getting the improvement in performance of image retrieval is the
theme of the work presented here. Many current CBIR systems use Euclidean distance [1-3,8-14]
on the extracted feature set as a similarity measure. The Direct Euclidian Distance between image
P and query image Q can be given as equation 1, where Vpi and Vqi are the feature vectors of
image P and Query image Q respectively with size ‘n’.
n
ED
2
=
∑ V
( pi − Vqi)
(1)
i=1

Total seven well-known image transforms [10,11,18,28] like Discrete Cosine Transform (DCT),
Walsh Transform, Haar Transform, Kekre Transform, Discrete Sine Transform (DST), Slant
Transform and Hartley Transform are used for performance comparison of the proposed CBIR
techniques.


FIGURE 1: Feature Extraction in Proposed CBIR Technique with Row Mean of Transformed Image
Columns

2. CBIR USING ROW MEAN OF TRANSFORMED COLUMN IMAGE [28]
Here image transform is applied on each column of image. Then row mean of the transformed
columns is used as feature vector. Figure 1 shows the Feature Extraction in Proposed CBIR
Technique with Row Mean of Transformed Image Columns. The obtained feature vector is used in
two different ways (with and without DC component) to see the variations in retrieval accuracy. As
indicated by experimental results, image retrieval using DC component value proves to be better
than retrieval excluding it.
Advances in Multimedia - An International Journal (AMIJ), Volume (2) : Issue (1) : 2011
2

---

Dr. H. B. Kekre, Sudeep D. Thepade & Akshay Maloo
The following steps need to be followed for image retrieval using the proposed image retrieval
techniques:
1. Apply transform T on the column of image of size NxN (INxN) to get column transformed
image of the same size (cINxN)
cINxN (column transformed) = [TNxN] [INxN]


(2)
2. Calculate row mean of column transformed image to get feature vector of size N (instead of
N2)
3. The feature vector is considered with and without DC component to see variations in results.
Then Euclidean Distance is applied to obtain precision and recall.
Applying transform on image columns instead of applying transform on the whole image, saves
50% of computations required resulting in faster retrieval [28]. Again row mean of column
transformed image is taken as feature vector which further reduces the required number of
comparisons among feature vectors resulting in faster retrieval. The results obtained from
proposed techniques of row mean of column transformed image with DC component and row
mean of column transformed image without DC component are compared with applying transform
on full image and spatial row mean of image in both gray and color versions of image database.
3. IMPLEMENTATION
3.1 The Platform and Image Database

The implementation of the proposed CBIR techniques is done in MATLAB 7.0 using a computer
with Intel Core 2 Duo Processor T8100 (2.1GHz) and 2 GB RAM.
The proposed CBIR techniques are tested on the image database of 1000 variable size images
spread across 11 categories of human being, animals, natural scenery and manmade things. This
image database in augmented version of Wang image database [15]. Figure 2 shows sample
image of generic database.



FIGURE 2: Sample Images from Generic Image Database
[Image database contains total 1000 images with 11 categories]



Advances in Multimedia - An International Journal (AMIJ), Volume (2) : Issue (1) : 2011
3

---

Dr. H. B. Kekre, Sudeep D. Thepade & Akshay Maloo
3.2 Precision/Recall
To assess the retrieval effectiveness, we have used the precision and recall as statistical
comparison parameters [1,2] for the proposed CBIR techniques. The standard definitions for
these two measures are given by following equations.

Number _ of _ relevant _ images _ retrieved
Pr ecision =

(3)
Total _ number _ of _ images _ retrieved

Number _ of _ relevant _ images _ retrieved
Re call

=
(4)
Total _ number _ of _ relevent _ images _ in _ database
4. RESULTS AND DISCUSSIONS
For testing the performance of each proposed CBIR technique, per technique 55 queries (5 from
each category) are fired on the database of 1000 variable size generic images spread across 11
categories. The query and database image matching is done using Euclidian distance. The
average precision and average recall values for each proposed technique with respective image
transform are computed and plotted against number of retrieved images for performance
comparison.
The crossover point of precision and recall plays very important role in performance analysis of
image retrieval method. At this crossover point value of precision equals to that of recall, which
means all the relevant images from database have been retrieved and are exactly equal to the
number of retrieved result images. In ideal situation the height of precision and recall crossover
point should be at value one, which means all the retrieved images are relevant and all relevant
from database are retrieved. Always the performance of image retrieval technique is compared to
this ideal situation. The height of crossover point of precision and recall gives idea about how
much the proposed technique is deviating from ideal one, more the height better the technique is.
The performance of proposed techniques with DC component (referred as ‘Transform-RM-DC’)
and without DC component (referred as ‘Transform-RM’) for each transform is compared with
CBIR using complete transformed image as feature vector (referred as ‘Full’), spatial row mean
vector of image as feature vector (referred as ‘RM’).
The proposed techniques are tested for both grey and color versions of image database. In all
image transforms the color versions of the discussed CBIR techniques give higher performance as
compared to gray versions.

4.1 Results on Gray Version of Image Database
In Figure 3 the precision-recall crossover points of DCT applied to the full gray image (Full), gray
row mean (RM), the proposed technique of row mean of DCT transformed gray columns applied
with DC component (DCT-RM-DC) and without DC component (DCT-RM) are shown. Here the
proposed method with DC component gives the highest crossover point indicating best
performance. Even the computational complexity in proposed retrieval technique is less than that
of applying full transform. This proves proposed image retrieval method is faster and better with
DCT. The performance of proposed CBIR method degrades if the DC component is not
considered.
Advances in Multimedia - An International Journal (AMIJ), Volume (2) : Issue (1) : 2011
4

---

Dr. H. B. Kekre, Sudeep D. Thepade & Akshay Maloo


FIGURE 3: Gray Crossover Point of Precision and Recall v/s Number of Retrieved Images using DCT




FIGURE 4: Gray Crossover Point of Precision and Recall v/s Number of Retrieved Images using DST

Advances in Multimedia - An International Journal (AMIJ), Volume (2) : Issue (1) : 2011
5

---