Journal of Software Engineering and Applications, 2013, 6, 131-139
http://dx.doi.org/10.4236/jsea.2013.63017 Published Online March 2013 (http://www.scirp.org/journal/jsea)
131
An Alternate Approach for Designing a Domain Specific
Image Search Prototype Using Histogram
Sukanta Sinha1, Rana Dattagupta2, Debajy oti Mukhopad hyay1,3
1Web Intelligence and Distributed Computing Research Lab, Kolkata, India; 2Computer Science Department, Jadavpur University,
Kolkata, India; 3Information Technology Department, Maharashtra Institute of Technology, Pune, India.
Received January 31st, 2013; revised February 28th, 2013; accepted March 8th, 2013
ABSTRACT
Everyone knows that thousand of words are represented by a single image. As a result, image search has become a very
popular mechanism for the Web-searchers. Image search means, the search results are produced by the search engine
should be a set of images along with their Web-page Unified Resource Locator (URL). Now Web-searcher can perform
two types of image search, they are “Text to Image” and “Image to Image” search. In “Text to Image” search, search
query should be a text. Based on the input text data, system will generate a set of images along with their Web-page
URL as an output. On the other hand, in “Image to Image” search, search query should be an image and based on this
image, system will generate a set of images along with their Web-page URL as an output. According to the current sce-
narios, “Text to Image” search mechanism always not returns perfect result. It matches the text data and then displays
the corresponding images as an output, which is not always perfect. To resolve this problem, Web researchers have in-
troduced the “Image to Image” search mechanism. In this paper, we have also proposed an alternate approach of “Image
to Image” search mechanism using Histogram.
Keywords: Domain Specific Crawling; Histogram; Gray-Scale Image; Image Search; Ontology; Search Engine
1. Introduction
Nowadays people are uploading the number of images in
the internet [1-4]. As a result internet has become a huge
reservoir of digital images. Exponentially increasing
digital image database volume has urged many research-
ers for developing effective image retrieval methods.
Considering the importance of the problem, various re-
searches have been carried out on image search over the
past few years and all are available in the literature.
There are two types of image search available such as
“Text to Image” search and “Image to Image” search.
The “Text to Image” search mechanism expects a search
text as a search query and following that search text, the
search prototype will generate a set of images. “Text to
Image” search approach, mainly matches the image tag
information like image name, meta-tag information, etc.
Now, consider such a situation where users upload their
images with irrelevant text information. That time this
approach will not work properly. In “Image to Image”
search mechanism, search query itself contains the search
image and based on the search image, search prototype
will generate a set of images as an output. Unlike re-
trieval of “Text to Image”, image search is difficult and
has involved image analysis. In this paper, an attempt has
been made to design a methodology for a domain spe-
cific image search prototype using histogram. This pro-
totype fully deals with “Image to Image” search and it is
a domain specific approach.
The paper is organized in the following way. In Sec-
tion 2, the basic of histogram and gray-scale of an image
are discussed. Section 3 discusses the existing works.
The proposed architecture for domain specific image
search prototype using histogram is discussed in Section
4. All the components of our architecture have been also
discussed in the same section. The experimental analysis
and the conclusions reached from our study have been
summarized in Sections 5 and 6.
2. Histogram Basics
Image histograms are an important concept in Image
Processing [5-9]. The histogram of an image refers to the
histogram of the intensity values of the pixels. Histogram
displays the number of pixels in an image for a particular
intensity level. A histogram is a graphical representation
of date distribution. An image histogram is a type of his-
togram that represents a digital image tonal data distribu-
tion. It plots the number of pixels for each tonal value.
By looking at the histogram for a specific image a viewer
will be able to judge the entire tonal distribution at a
glance.

An Alternate Approach for Designing a Domain Specific Image Search Prototype Using Histogram
132
2.1. Why Choose Histogram
Web-pages have contained various types of images. It
may be color image, gray-scale image, black-and-white
image or various types of image formats like .jpg, .jpeg,
etc. To perform an image search, we need to make all the
images in a common format. The histogram is one of the
simple and useful tools to process those images and pro-
duce a common format.
2.2. What Do You Mean by Gray-Scale Image
A gray-scale digital image is an image in which the value
of each pixel is a single sample, that is, it carries only
intensity information. Images of this sort, also known as
black-and-white, are composed exclusively of shades of
gray, varying from black at the weakest intensity to white
at the strongest [10].
2.3. Histogram of a Gray-Scale Image
The gray-scale histogram of an image represents the dis-
tribution of the pixels in the image over the gray-level
scale. It can be visualized as if each pixel is placed in a
bin corresponding to the color intensity of that pixel. All
of the pixels in each bin are then added up and displayed
on a graph. This graph is the histogram of the image.
3. Existing Work
Image search is such a complex mechanism, where vari-
ous researches are going on to improve the search proto-
type. Our paper is not intended to provide a complete
survey of techniques. According to our knowledge, we
have applied these techniques on few examples. Now a
day’s research on search engine has been carried out in
universities and open laboratories, many dot-com com-
panies. Unfortunately, many of these techniques are used
by dot-coms, and especially the resulting performance,
are kept private behind company walls, or are disclosed
in patents that can be comprehended and appreciate by
the lawyers. Therefore, we believe that the overview of
problems and techniques that we presented here can be
useful.
In this section, we have explained few existing mecha-
nisms and explained how the current systems are work-
ing.
3.1. Definitions
 Ontology—It is a set of domain related key informa-
tion, which is kept in an organized way based on their
importance.
 Relevance Value—It is a numeric value for each
Web-page, which is generated on the basis of the term
Weight value, term Synonyms, number of occurren-
ces of Ontology terms which are existing in that Web-
page.
 Seed URL—It is a set of base URL from where the
crawler starts to crawl down the Web pages from the
Internet.
 Weight Table—This table has two columns, first
column denotes Ontology terms and second column
denotes weight value of that Ontology term. Ontology
term weight value lies between “0” and “1”.
 Syntable—This table has two columns, first column
denotes Ontology terms and second column denotes
synonym of that ontology term. For a particular on-
tology term, if more than one synonym exists, those
are kept using comma (,) separator.
 Relevance Limit—It is a predefined static relevance
cut-off value to recognize whether a Web-page is
domain specific or not.
 Term Relevance Value—It is a numeric value for
each Ontology term, which is generated on the basis
of the term Weight value, term Synonyms, number of
occurrences of that Ontology term in the considered
Web-page.
3.2. Domain Specific Crawling
Domain specific crawling means the Web crawler crawls
only domain specific Web-pages [11-18]. For finding
domains based on the Web-page content, first parsed the
Web-page content and then extracted all the Ontology
terms as well as syntable terms [19-22]. Then each dis-
tinct Ontology term was multiplied with their respective
Ontology term weight value. Ontology term weight val-
ues are taken from weight table. In this approach, for any
syntable term used corresponding Ontology term weight
value. Finally, taken a summation of these individual
terms weigh value and this value is called relevance
value of that Web-page.
Now if this relevance value is greater than the prede-
fined “Relevance Limit” of that domain, then that Web-
page belongs to a predefined particular domain otherwise
discard the Web-page, i.e., the Web-page didn’t belong
to our domain. In Figure 1 we have shown a mechanism
Figure 1. Web-page relevance cal cu la ti on mechanism.

An Alternate Approach for Designing a Domain Specific Image Search Prototype Using Histogram
133
to find a domain based on the Web-page content. Here,
we consider “computer science” Ontology, syntable and
weight table of computer science Ontology for finding a
Web-page belongs to the computer science domain or not.
Suppose, the considered Web-page contains “student”
term 3 times, “lecturer” term 2 times and “associate pro-
fessor” term 2 times and student, lecturer and associate
professor weight values in the computer science domain
are 0.4, 0.8 and 1.0 respectively. Then the relevance
value becomes (3 × 0.4 + 2 × 0.8 + 2 × 1.0) = 4.8. Now,
if 4.8 is greater than the relevance limit, then we called
the considered Web-page belongs to the computer sci-
ence domain otherwise we discard the Web-page.
not “A” user image. Hence it is an invalid search result
for “B” user’s side. To resolve this problem we need to
analyze the images and then only we can produce correct
search results. For that reason Web researcher are intro-
duced “Image to Image” search mechanism which ex-
plained in the next subsection.
3.4. Existing Image to Image Search
In “Image to Image” search Web searcher gives an image
as a search query [27-29]. Now there are various types of
images such as color image, gray-scale image, black-and-
white image and different image file formats like .jpg,
.jpeg, .bmp, .tif, etc. are available in the internet. We
have done a survey where we found some issues. For
popular images like “Rabindranath Tagore”, “Sachin
Tendulkar”, etc., the search engines are working fine
(refer Figure 2) but those images which are not a popular
image that time we have received a lot of irrelevant re-
sults. Suppose we have a “xyz” image, which is not a
popular image but available in few websites like “face-
book”, “LinkedIn”, etc. Using this “xyz” image while we
performing the search operation, we have not received
relevant results (refer Figure 3). We have also found
another issue, say I have a color image and lots of
Web-pages are exists in the internet which holding the
3.3. Existing Text to Image Search
In “Text to Image” search mechanism Web searcher will
provide a search text and based on that text system will
find the images [23-26]. Let me explain by taking an
example where this mechanism will not work. Consider a
Web user (A) was created a “facebook” profile using
Sachin Tendulkar’s image, which is an invalid image
with respect to the user (A). Now another Web user (B)
wants to see the “A” user image that time if “B” user
performs a text to image search based on A’s user name.
The search result will produce Sachin Tendulkar’s image
Figure 2. Image search for a popular image (Sachin Te ndulkar ) .

An Alternate Approach for Designing a Domain Specific Image Search Prototype Using Histogram
134
Figure 3. Image search for a non-popular image (xyz).
same image but they are black-and-white image. In this
scenario, if we put a color image as a search image and
there does not exist any Web-page which contains the
same color image that time also we found a lot of irrele-
vant results. As a result the Web users were misguided.
To resolve this problem we have proposed an alternate
image search method using histogram which can be use-
ful and described in successive sections.
4. Our Approach
There are various approaches followed by the Web re-
searchers for searching an image from the internet. We
also proposed a unique alternate method for image
searching and we believe that this approach can be useful.
Dealing with images is always a huge challenge. There
are two major difficulties we have faced. In the internet
lot of images are exists and there we have found some
similar images with different size. Those images are pre-
sent in different Web-pages, i.e., Web-page URLs are
different. In that case we have used % of pixel distribu-
tion to resolve the various types of image size used. The
“percentage (%) of pixels distribution” calculation logic
has given in Equation (1).
Percentage (%) of pixels exist in xi th position
  
TNP 100%
i
Px   (1)
where, 0 ≤ i ≤ 255; TNP = Total number of pixels exist
in the image; λ = Number of pixels found in xi th position
of the histogram.
Internet is a conglomeration of huge amount of multi-
colored images. To avoid the color variations, we have
converted all the images into 8-bit gray-scale image. In
our approach, we have divided our proposed process into
small modules like “generation of image repository”,
“Search result generation”, etc. In subsequent sections,
we have explained these modules.
4.1. Image Repository Creation
Image repository creation is an important role to perform
“Image to Image” search. Our proposed algorithm has
given below:
Input: Seed URLs, Weight Table, Syntable
Output: Image repository
Step 1: Initialize the crawlers by the Seed URLs.
Step 2: Crawl the Web-pages from internet.
Step 3: Calculate relevance value.
Step 4: If Relevance Value > Relevance Limit then
If Web-page contains any image then
1) Get the image and save the image.
2) Convert the image into 8-bit gray scale image.
3) Generate histogram of that gray scale image.
4) Calculate percentage (%) of pixels distribution us-
ing Equation (1).
5) Save all the % of pixels distribution.
Step 5: Go to step-2 until crawler has no URL.
Step 6: End.
In Figure 4, we have shown the image repository
creation mechanism. Before running the crawler, we

An Alternate Approach for Designing a Domain Specific Image Search Prototype Using Histogram 135
Figure 4. Image repository generation.
need to set the seed URLs, weight table and syntable. We
have used MATLAB 7.10 for generating histograms.
First crawler crawls the domain specific Web-pages from
the internet. Then identify the Web-pages which contain
images and get the images for further processing. Con-
nect the images into 8-bit gray-scale image and generate
histograms. Now using Equation (1) calculates percent-
age of pixel distributions and save it for producing search
results.
4.2. Search Result Generation Mechanism
To produce search results, we need to analyze the search
image and then matched with our repository image at-
tributes. We are basically matching all P(xi) where “i”
belongs to [0, 255] and the matching dine between search
image and our repository image (refer Figure 5). Now
the basic concept of our proposed search result genera-
tion mechanism is divided into two parts one exact match
images and probable matched images. In exact match
search mechanism, users will receive those Web-pages as
a search result where search image was exactly matched
with repository images except their size and color, be-
cause we are proposing such a prototype where image
size and color has no impact on producing search results.
In probable match search mechanism, we have used a
tolerance value to match the percentage (%) of distribu-
tions and then produced the search results. Tolerance
Figure 5. Proposed search mechanism.
value is a user given number. For exact match case it is
zero (0) and other cases it must be a positive integer
number in the range of [0, 100].
4.3. User Interface
Figure 6 shows a part of the user interface of our im-
age search prototype. An image URL is typed or browses
in the input image URL box, select either exact image or
probable image match radio button, enter tolerance value,
select the domain and enter the relevance range. In the
user interface mandatory fields are denoted by star (*)
sign. Relevance range default value set as [maximum
relevance value, minimum relevance value], which is an
editable field and according to the requirement, user can
customize the relevance range values. We are providing
flexibility to the users to get exact matched images or
probable matched images by selecting the radio buttons
in the user interface. While we select exact match that
time tolerance value field become read only and set as
zero (0) on the other hand if user selects probable image
match radio button in the user interface that time toler-
ance value field becomes editable and defaulted as zero
(0). In the user interface, the maximum relevance value
and minimum relevance value are set dynamically ac-
cording to the user selected domain. While refreshing the
database, maximum relevance value is taken using ceil-
ing function for the largest Web-page’s relevance value
and minimum relevance value is taken using floor func-
tion for the smallest Web-page’s relevance value. Here,
three domains such as “Cricket”, “Hockey”, and “Foot
Ball” have been considered.

An Alternate Approach for Designing a Domain Specific Image Search Prototype Using Histogram
136
Figure 6. A part of user interface.
For domain selection, we have used radio button be-
cause at a time only one domain can be selected. User
first input all the necessary inputs and then click on
“Search” button of the user interface, that time our pro-
posed search result generation mechanism called to pro-
duce the search results.
5. Experimental Analysis
In this section we have shown our test settings, how the
images were analyzed and some sample image search
results.
5.1. Test Settings
In this subsection we will describe different parameter
settings to crawl domain specific Web-pages and their
images. To run the crawler, we need to set the seed URLs,
weight table and syntable. We have shown few sample
data in Tables 1-3.
5.2. Image Analysis
In this subsection, we have shown the image analysis and
given a partial percentage of the pixels distribution chart.
Our main aim is, for a single image, color and size do not
affect on image searching. We have shown six images in
Figure 7(a), (c), (e), (g), (i), (k). We have considered
those six figures such a way where, two same color im-
ages with different size, two same black & white images
with different size and two gray scale images with dif-
ferent size. Now from those images we have generated
histogram (refer Figure 7(b), (d), (f), (h), (j), (l)) and
then using Equation (1), we have generated in Table 4.
From the table, we have seen color and size doesn’t mat-
ter because P(xi) are holding approximately same value.
5.3. Image Search Results
We have used a domain specific approach, so that lots of
Table 1. Seed URLs.
Seed URLs
http://icc-cricket.yahoo.com/
http://www.cricketnext.com/index.html
http://www.in.com
http://www.cricketworld.com/
Table 2. Weight Table.
Ontology terms Weight value
cricket 0.9
wicket keeper 0.8
umpire 0.4
bat 0.2
match 0.1
Table 3. Syntable.
Ontology terms Synterms
match competition, contest
stamp stick, wicket
ball conglobate, conglomerate
umpire judge, moderator, referee
catch capture
unwanted images are already eliminated from the re-
pository in crawling phase. As discussed in Section 3.4,
we have faced some issues while performing an image
search in existing search prototype. Same issues we can
resolve using our prototype. We have considered same
search image and produced the search results. In Figure
8, we have shown our prototype search results. From user
interface we have selected the exact match option while
performed the search operation.
6. Conclusion
In this paper, we have proposed an alternate approach for
designing a domain specific image search prototype us-
ing the histogram. Our prototype is mainly designed for
few domains. Moreover, this prototype is highly scalable.
We can expand supporting domains by introducing new
domain Ontology and other details such as weight table,
syntable, etc. Our prototype image search doesn’t depend
on search image color and size. We have created a uni-
form color and size image repository using histogram
and Equation (1) percentage of pixel distribution mecha-
nism. We have facilitated exact match and probable
match option to the Web searchers using tolerance value.

An Alternate Approach for Designing a Domain Specific Image Search Prototype Using Histogram
137
Figure 7. (a) 256 × 256 color image; (b) 256 × 256 color image histogram after gray-scale conversion; (c) 256 × 256 gray-scale
image; (d) 256 × 256 gray-scale image histogram; (e) 256 × 256 black-and-white image; (f) 256 × 256 black-and-white image
histogram; (g) 128 × 128 color image; (h) 128 × 128 color image histogram after gray-scale conversion; (i) 128 × 128
gray-scale image; (j) 128 × 128 gray-scale image histogram; (k) 128 × 128 black-and-white image; (l) 128 × 128 black-
and-white image histogram.

An Alternate Approach for Designing a Domain Specific Image Search Prototype Using Histogram
138
Table 4. Percentage (%) of pixels distribution chart.
Percentage (%) of pixels exist in various xi th position
Histogram
Reference Figure P(x25) P(x50) P(x75) P(x100) P(x125) P(x150) P(x175) P(x200) P(x225) P(x250)
Figure 7(b) 0.435 0.427 0.443 0.473 0.443 0.428 0.366 0.344 0.435 0.205
Figure 7(d) 0.435 0.427 0.443 0.473 0.443 0.428 0.366 0.344 0.435 0.205
Figure 7(f) 0.434 0.428 0.443 0.473 0.444 0.427 0.365 0.343 0.436 0.206
Figure 7(h) 0.436 0.427 0.444 0.472 0.443 0.427 0.366 0.343 0.436 0.206
Figure 7(j) 0.436 0.427 0.444 0.472 0.443 0.427 0.366 0.343 0.436 0.206
Figure 7(l) 0.433 0.427 0.443 0.473 0.442 0.426 0.366 0.345 0.435 0.205
Search Image (You have selected exact match option)Search Image (You have selected exact match option)
Search Results Search Results
http://www.facebook.com/
http://www.in.com/
http://www.in.com/sachin-tendulkar/profile-50.html
http://www.biography.com/people/sachin-tendulka-9503921
(a)
(
b
)
Figure 8. Image search result: (a) Non-popular image; (b) Popular image.
We are giving a wide area to the Web searchers by taking
tolerance value from the user interface. While doing the
survey of existing works, we found there are lot of “Im-
age to Image” searches are going on behind the dot-com
companies. But still we have found some issues on image
to image search, which we can resolve using our proto-
type.
REFERENCES
[1] T. Berners-Lee and M. Fischetti, “Weaving the Web: The
Original Design and Ultimate Destiny of the World Wide
Web by Its Inventor,” HarperBusiness, New York, 1999.
[2] W. Willinger, R. Govindan, S. Jamin, V. Paxson and S.
Shenker, “Scaling Phenomena in the Internet,” Proceed-
ings of the National Academy of Sciences, New York, 19
February 2002, pp. 2573-2580.
[3] J. J. Rehmeyer, “Mapping a Medusa: The Internet Spreads
Its Tentacles,” Science News, Vol. 171, No. 25, 2007, pp.
387-388. doi:10.1002/scin.2007.5591712503
[4] B. M. Leiner, V. G. Cerf, D. D. Clark, R. E. Kahn, L.
Kleinrock, D. C. Lynch, J. Postel, L. G. Roberts and S.
Wolff, “A Brief History of Internet,” ACM Computer
Communication, Vol. 35, No. 1, 2009, pp. 22-31.
doi:10.1145/1629607.1629613
[5] X. Wang, J. Wu and H. Yang, “Robust Image Retrieval
Based on Color Histogram of Local Feature Regions,”
Springer, Berlin, 2009.
[6] C. L. Novak and S. A. Shafer, “Anatomy of a Color His-
togram,” Proceedings of IEEE Computer Society Confe-
rence on Computer Vision and Pattern Recognition, Ch-
ampaign, 15-18 June 1992, pp. 599-605.
doi:10.1109/CVPR.1992.223129
[7] J. R. Smith, “Integrated Spatial and Feature Image Sys-
tems: Retrieval, Analysis and Compression,” Ph.D. The-
sis, Columbia University, New York, 1997.

An Alternate Approach for Designing a Domain Specific Image Search Prototype Using Histogram 139
[8] E. A. Bashkov and N. S. Kostyukova, “Effectiveness
Estimation of Image Retrieval by 2D Color Histogram,”
Journal of Automation and Information Sciences, Vol. 8,
No. 11, 2006, pp. 74-80.
[9] E. A. Bashkov and N. S. Shozda, “Content-Based Image
Retrieval Using Color Histogram Correlation,” Graphi-
con proceedings, Nizhny Novgorod, 16-21 September
2002, pp. 458-461.
[10] L. Vincent, “Morphological Grayscale Reconstruction in
Image Analysis: Applications and Efficient Algorithms,”
IEEE Transactions on Image Processing, Vol. 2, No. 2,
1993, pp. 176-201.
[11] D. Mukhopadhyay, A. Biswas and S. Sinha, “A New
Approach to Design Domain Specific Ontology Based
Web Crawler,” The 10th International Conference on In-
formation Technology, Orissa, 17-20 December 2007, pp.
289-291.
[12] S. Chakrabarti, M. Berg and B. E. Dom, “Focused
Crawling: A New Approach to Topic-Specific Web Re-
source Discovery,” Proceedings of the 8th International
World Wide Web Conference, Toronto, 11-14 May 1999,
pp. 545-562.
[13] D. Bergmark, C. Lagoze and A. Sbityakov, “Focused
Crawls, Tunneling, and Digital Libraries,” Proceedings of
the European Conference on Digital Libraries, Vol. 2458,
Rome, 16-18 September 2002, pp. 91-106.
[14] M. Diligenti, F. Coetzee, S. Lawrence, C. L. Giles and M.
Gori, “Focused Crawling Using Context Graphs,” The
26th International Conference on Very Large Databases,
Cairo, 10-14 September 2000, pp. 527-534.
[15] N. Tyagi and D. Gupta, “A Novel Architecture for Do-
main Specific Parallel Crawler,” Indian Journal of Com-
puter Science and Engineering, Vol. 1 No. 1, 2008, pp.
44-53.
[16] A. Kundu, R. D. Dattagupta and D. Mukhopadhyay,
“Mining the Web with Hierarchical Crawlers—A Re-
source Sharing Based Crawling Approach,” International
Journal on Intelligent Information and Database Systems,
Vol. 3, No. 1, 2009, pp. 90-106.
doi:10.1504/IJIIDS.2009.023040
[17] A. Kundu, R. Dutta, D. Mukhopadhyay and Y. C. Kim,
“A Hierarchical Web Page Crawler for Crawling the In-
ternet Faster,” Proceedings of the International Confer-
ence on Electronics and Information Technology Con-
vergence, Republic of Korea, 20 December 2006, pp.
61-67.
[18] D. Mukhopadhyay, S. Mukherjee, S. Ghosh, S. Kar and Y.
C. Kim, “Architecture of a Scalable Dynamic Parallel
Web Crawler with High Speed Downloadable Capability
for a Web Search Engine,” Proceedings of the 6th Inter-
national Workshop, Republic of Korea, 20 November
2006, pp. 103-108.
[19] A. Gangemi, R. Navigli and P. Velardi, “The Onto-
WordNet Project: Extension and Axiomatization of Con-
ceptual Relations in WordNet,” Proceedings of Interna-
tional Conference on Ontologies, Databases and Appli-
cations of Semantics, Catania, 3-7 November 2003, pp.
820-838.
[20] D. N. Antonio, M. Michele and N. Roberto, “A Software
Engineering Approach to Ontology Building,” Informa-
tion Systems, Vol. 34, No. 2, 2009, pp. 258-275.
doi:10.1016/j.is.2008.07.002
[21] T. Gruber, “Toward Principles for the Design of Ontolo-
gies Used for Knowledge Sharing,” International Journal
of Human-Computer Studies, Vol. 43, No. 5-6, 1995, pp.
907-928. doi:10.1006/ijhc.1995.1081
[22] M. Ehrig and A. Maedche, “Ontology-Focused Crawling
of Web Documents,” Proceedings of the ACM Sympo-
sium on Applied Computing, Melbourne, 9-12 March
2003, pp. 1174-1178.
[23] B. Luo, X. G. Wang and X. O. Tang, “World Wide Web
Based Image Search Engine Using Text and Image Con-
tent Features,” Proceedings of SPIE Electronic Imaging,
Vol. 5018, Santa Clara, 20 January 2003, pp. 123-130.
[24] Z. Su, H. Zhang, S. Li and S. Ma, “Relevance Feedback
Content-Based Image Retrieval: Bayesian Framework,
Feature,” IEEE Transactions on Image Processing, Vol.
12, No. 8, 2003, pp. 924-937.
[25] K. P. Yee, K. Swearingen, K. Li and M. Hearst, “Faceted
Metadata for Image Search and Browsing,” Proceedings
of the ACM Digital Library, Fort Lauderdale, 5-10 April
2003, pp. 401-408.
[26] C. W. Niblack, R. Barber, W. Equitz, M. Flickner, E.
Glasman, D. Pektovic, P. Yanker, C. Faloutsos and G.
Taubin, “The QBIC Project: Querying Images by Content
Using Color, Texture, and Shape,” Proceedings of Stor-
age and Retrieval for Image and Video Databases, San
Jose, 31 January 1993, pp. 173-187.
[27] D. N. D. Harini and D. L. Bhaskari, “Image Retrieval
System Based on Feature Extraction and Relevance Feed-
back,” Proceedings of the CUBE International Informa-
tion Technology Conference, Pune, 3-5 September 2012,
pp. 69-73.
[28] A. K. Jain and A. Vailaya, “Image Retrieval Using Color
and Shape,” Pattern Recognition, Vol. 29, No. 8, 1996,
pp. 1233-1244.
[29] D. Patra and J. Mridula, “Featured Based Segmentation of
Color Textured Images Using GLCM and Markov Ran-
dom Field Model,” World Academy of Science, Engi-
neering and Technology, Vol. 53, No. 5, 2011, pp. 108-
113.