Content Based Search System for Ethiopian Art Content on the Web

Hewan Shilema Mesfin Belachew

HiLCoE, Software Engineering Programme, Ethiopia HiLCoE, Ethiopia
[email protected] Ministry of Communication and Information
Technology
[email protected] 

Abstract
In this paper a content based search system is designed and implemented to allow users to access different
art works of Ethiopian artists and makes searching information from the Web very easy and quick. Search
engines were originally described as automated programs that compile and update databases without human
intervention to serve as a bridge between a user and the artworks. Our search engine is domain-specific.
Domain-specific Web search engines are effective tools for reducing the difficulty experienced when looking
for information from the Web. Existing methods for building domain-specific Web search engines require
human expertise or specific facilities. However, we can build a domain-specific search engine simply by adding
domain-specific keywords to the user's input query.
In this paper, the necessity of designing an efficient search system and the significance of this work is
addressed in considering the given drawbacks in the existing management systems for the available art works.
Then an overview of current image database search methods is presented, followed by explanation of the
method that was selected for the construction of our content based search system and the reasons for choosing
the method. After this, the structure of the proposed contented based search system and its implementation are
described in detail.
Keywords: Search Engine; Content- Based Retrieval; Image Database

1. Introduction This characterization of CBIR as a field of study

places it at a unique point in time within the scientific
In many areas of commerce, government,
community. While we witness continued effort in
academia, and hospitals, large collections of digital
solving the fundamental open problem of robust
images are being created. Many of these collections
image understanding, a lateral bridging of gaps
are the product of digitizing existing collections of
between some of these research communities is being
analog photographs, diagrams, drawings, paintings,
gradually brought about as a by-product of such
and prints. Usually, the only way of searching these
contributions, the impact of which can potentially go
collections was by keyword indexing, or simply by
beyond CBIR. Again, what we see today as a few
browsing. Digital image databases, however, open
cross-field publications may very well spring into
the way for content-based searching [3].
new fields of study in the foreseeable future [5].
Content-Based Image Retrieval (CBIR), as we see
Nowadays there are a number of art galleries in
it today, is a technology that in principle helps to
Ethiopia and these galleries have their own websites.
organize digital picture archives by their visual
Art gallery websites display different art works that
content. By this definition, anything ranging from an
they show on the exhibit time and famous painters’
image similarity function to a robust image
works are found on different art galleries. From this
annotation engine falls under the purview of CBIR.
anyone who loves to see or buy art works cannot find
16 Content Based Search System for Ethiopian Art Content on the Web

art works in a collected way. So having a search interests. Images were first annotated with text and
system minimizes the difficulty of getting art works then searched using a text-based approach from
from different art gallery sites and it combines all traditional database management systems.
artistic works together. As the diversity and size of digital image
We developed a system called content based collections have grown exponentially, efficient image
search system for Ethiopian art content on the web. retrieval is becoming increasingly important. Large
There are many ways available to gather information image databases are difficult to browse with
about art works, such as from newspapers, by asking traditional text searches because the task of user
information from friends, and online websites. Yet, based annotation become very time consuming, as
the information available in newspapers also may be the text often fails to convey the rich structure of
mostly from large companies/galleries advertising images. A content-based retrieval system solves this
their newly shows. Information obtained from friends problem where retrieval is based on the automating
may be limited or similar as they may obtain them matching of features of the query image with that of
from the same source. image database through some image-image similarity
In order to improve the relevance and usefulness evaluation. Therefore, images will be indexed
of the search for art work information for a particular according to their visual content such as color,
location and therefore to reduce the time the user texture, shape or any other feature or a combination
would need to spend in analyzing the search result, of visual features [8].
content based search system for Ethiopian art content Content-based image retrieval uses the visual
on the web is developed. It enables and works as a contents of an image such as color, shape, texture,
bridge between a user and the artworks. Search and spatial layout to represent and index the image.
engines were originally described as automated In typical content-based image retrieval systems, the
programs that compiled and updated databases visual contents of the images in the database are
without human intervention. extracted and described by multi-dimensional feature
Using a search engine, the problems incurred in vectors [7].
the previous artwork systems can be resolved. Using 2.2 Content-Based Retrieval Systems
efficient and advanced compression techniques, we Content-based retrieval uses the contents of
can digitize a large image effectively. images to represent and access the images. A typical
Therefore, a large piece of artwork can be easily content-based retrieval system is divided into off-line
presented to users. By providing a reasonable feature extraction and online image retrieval [1]. A
digitized copy of the original artworks, we avoid conceptual framework for content-based image
directly handling some delicate artworks. retrieval is illustrated in Figure 1 [1]. In offline
feature extraction, the contents of the images in the
2. Related Work
database are extracted and described with a multi-
2.1 Overview of Content-Based Image Retrieval dimensional feature vector, also called descriptor.
One of the tools that will be essential for future The feature vectors of the image constitute a feature
electronic publishing is a powerful image retrieval dataset stored in a database. In online image retrieval,
system. Image retrieval systems associate keywords the user can submit a query example to the retrieval
or text with each image and require the user to enter system in search of desired images. Then the system
a keyword or textual description of the desired image represents this example with a feature vector [1].
[6]. Content-based image retrieval is a technique
which uses visual contents to search images from
large scale image databases according to users'
HiLCoE Journal of Computer Science and Technology, Vol. 3, No. 2 17 

Figure 1: Conceptual Framework for CBIR

At the end, the system ranks the search results and 2.3 Content Based Image Retrieval Frameworks
then returns the results that are most similar to the The basic idea of CBIR is that a set of features is
query examples. If the user is not satisfied with the used that allows finding images that are similar to the
search results, the user can provide feedback to the query image. For different properties of images,
retrieval system, which contains a mechanism to different features may be used. The goal of feature
learn the user’s information needs. The following extraction is to find informative variables based on
sections introduce each component in the system [1]. image data which can be seen as a kind of data
Most CBIR systems have a single-interaction reduction [4].
modality where the user presents an image as a query The user interface normally consists of a query
and obtains matching images as results. This one- formulation part and a result presentation part.
shot modality for querying information repositories Specification of which images to retrieve from the
has a history of success ranging from decades old database can be done in many ways. One way is to
classical and textual database queries and browse through the database one by one; another way
information retrieval techniques to more recent web is to specify the image in terms of keywords or in
searching facilities. Though such one-shot query terms of image features that are extracted from the
modality has been effective in dealing with large image, such as a color histogram. Still another way is
form of text-based information repositories, we to provide an image or sketch from which features of
believe that the same modality is undesirable, the same type must be extracted as for the database
ineffective, and inefficient for a significant class of images in order to match these features [3].
image retrieval domains. Hence, interactive CBIR Relevance feedback is about providing positive or
systems are gaining popularity in the past few years negative feedback about the retrieval result so that
[2]. the system can refine the search. Other features that
A number of keyword-based general Web search are used to specify queries are color, texture, shape,
engines allow image media types. A number of other spatial layout, and face. Color features are often
general search engines are more specifically designed obtained directly from the pixel intensities. Although
for images. There are many special image collections a precise definition of texture has been allusive, the
on the Web that can be searched with alphanumerical notion of texture generally refers to the presence of a
keywords. [3] spatial pattern that has some properties of
homogeneity. In particular, the homogeneity cannot
18 Content Based Search System for Ethiopian Art Content on the Web

result from the presence of only a single color in the  We want to display multiple images on the
regions, but requires interaction of various colors [3]. screen. Therefore we decided to use thumbnail
Higher level features are increasingly more sized images. One big advantage of using
specific, and thus less widely used. However, faces thumbnail sized images is that the user gets a
are frequently present in pictures and relatively often better overview with more than just one
used as a feature [3]. image. Also no information is lost because the
user is still capable of understanding the
3. The Proposed Solution
image's contents.
Automating the existing system through  Another advantage is digital storage and
centralized database with web based search system is retrieval which is essential to World Wide
chosen to address current problems that most artists Web programs.
face. Some problems that appear on most of the art
 Online image view is done online which saves
work system are the following.
time of the buyers of the art and decreases the
 Time and place constraints: In the current art workload of the galleries.
work management systems, the artworks are
stored in art gallery rooms, which means that 4. System Architecture
users need to go to the gallery to view these The architecture of the system shows the general
artworks. As far as the user is concerned, this structure of the search system for retrieving
is inconvenient. information from the database server. Figure 2
 Difficult or impossible physical access: It will illustrates the architecture of the system. As shown in
be difficult or impossible to access some the figure, the user search query is sent to the web
artworks that may be spread across a large server and then the web server sends the query to the
physical area or composed of material that index server and then the index server refers the
may require delicate handling. database server to search the user query and finally
 Safety: Some artworks are very valuable and the database server returns the search result for the
precious. Exposing these artworks directly to end user.
users probably will result in wear, Figure 3 shows the user interface for content
contamination and even damage. Therefore, it based search system for Ethiopian art content on the
is not safe to expose some real artworks to the Web. Anyone who wants to use this system can get
public. many art works and detailed information about the
 Lack of Information: users have no way to artist, art news, the address of art museums in
view the gallery collection as a whole and Ethiopia, announcements that help artists and art
have little or no idea about what is contained work lovers about art work exhibitions and upcoming
in the collection. events, and the system has also links that help a user
who wants to buy an art work to get detailed
Automating the existing system outsmarts the
information about the art work. There is also a forum
other alternative. So we decided to develop the
link that helps the user to discuss issues by creating
system as web based application and it has the
user account.
following benefits.
HiLCoE Journal of Computer Science and Technology, Vol. 3, No. 2 19 

Figure 2: System Architecture for Content Based Search System

Figure 3: Front Main Menu

Figure 4 displays the interface for the auction site detailed information about each art works, the price,
that help the user to get detailed information about status of the art work if it is sold or not, the location
artworks, the auction site lists the artworks by where the art work is found and which bank is
category, and anyone who wants to buy an artworks preferable for the payment process and so other
can select from the listed category and can see information are displayed on this auction site.
20 Content Based Search System for Ethiopian Art Content on the Web

Figure 4: Main Menu for the Auction Site

5. Conclusion and Future Work University, Department of Computer Science,
Utrecht, the Netherlands.
Content based search system for Ethiopian art
content on the web is developed to provide more [4] Zainab InrahimAbood, Israa Jemeel Muhsin,
relevant search results for users by allowing the user and Nabeel Jameel Tawfiq, “Content-Based
to input information by specifying keywords on the Image Retrieval (CBIR) Using Hybrid
search system. Technique”, International Journal of Computer
Applications (0975-8887), Vol. 83, No 12,
Currently due to globalization the world is highly
December 2013.
becoming competitive, so organizations have to
direct their attention on using recent technology to be [5] Ritendra Datta, Dhiraj Joshi, Jia Li, and James
competitive. This can be real if they are able to use Z. Wang, “Image Retrieval: Ideas, Influences,
information communication technology (ICT) to and Trends of the New Age”, The Pennsylvania
successfully achieve their objectives. State University.

Since the system was a proof-of-concept system, [6] Yuri Lausberg, “A Content-Based Image Search
it is not fully functional and requires more Engine in Cyberspace,” Application-Oriented
development in the future. Computer Science Leiden Imaging and

References Multimedia Group, Department of Computer

Science, Leiden University, August 19, 1996.
[1] K. Vijay and R. Anitha, “A Content-Based
Approach to Image Database Retrieval”, MCA, [7] Fuhui Long, Hongjiang Zhang, and David
K. S. R. College of Technology, Tiruchengode. Dagan Feng, “Fundamentals of Content-Based
[2] Sabyasachi Saha and Sandi Sen, “Agent Based Image Retrieval”.
Framework for Content Based Image Retrieval”, [8] Gaurav Jaswal, Amit Kaul, and Rajan Parmar,
Department of Mathematics and Computer “Content Based Image Retrieval Using Color
Science, The University of Tulsa. Space Approaches”, International Journal of
[3] Remco C. Veltkamp, MirelaTanase, and Engineering and Advanced Technology (IJEAT)
Denielle Sent, “Features in Content-Based Vol. 2, Issue-1, October 2012.
Image Retrieval Systems: A Survey”, Utrecht