Content Based Image Retrieval
Content Based Image Retrieval
Volume 4, Issue 3, March – 2019 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Content Based Image Retrieval
Roshni Sahu1, Shital Zade2, Sonal Kamble3, Payal Bhagat4, Bhaghyashri Awachat5, Mayur Dakhane6
1,2,3,4,5,6
Computer Science and Engineering Department, Datta Meghe Institute of Engineering, Technology and Research,
Rashtrasant Tukadoji Maharaj Nagpur University, Maharashtra, India
Abstract:- To the explosive growth of digital image performing OCR on image we can extract the text and
there is need of efficient and effective technique to perform the copy this text and also modify this text.
image retrieval from database or from the internet.
Now day no one can have that much of time that they CBIR process is used to retrieve image from image
will check or visit each image and check, is it required .feature like colour, shape, texture.
image or not. So Present new technique of image
retrieval system which is based on the combination of In CBIR image retrieval by using colour feature use
both text and content based. The idea behind TBIR Histogram. for shape use kNN (k nearest neighbours)
(text-based image retrieval) is that it uses the title, tag, algorithm, and for texture also use kNN algorithm. In kNN
character which is present inside the image i.e. algorithm use Euclidean formula. By using this formula we
metadata using this feature retrieve image. In CBIR find image with their actual value and original value.[2]
(content-based image retrieval) image retrieving based
on image feature i.e. colour, texture, shape, etc. Using II. RELATED WORK
these two techniques we can extract image from their
feature. Main aim by using this type of system is to The previous system implement or design the
reduce the time when we retrieve one image from combination of text based and content based image
database. retrieval image feature. They use the Flickr sidebar to
perform search and use image URL as a query input. When
Keywords:- Image Retrieval, TBIR (Text-Based Image user perform text based image retrieval and enter any text
Retrieval), CBIR (Content-Based Image Retrieval), Color, to search or the input contain text based feature, then in the
Shape, Texture, Metadata. output inverted file is used for extract the text or the set of
image that contain the input text. But when user does not
I. INTRODUCTION enter any text for search then image is search in linear
fashion over all the collection of image. [5]
There is huge amount of image are creating in digital
life and all are stored in database or gallery. In most of area Proposed method to extract feature from image. For
this image are most important like academia, commerce, retrieving image they use following techniques.
government sector, medical or health care, and internet.
In this the texture classification they used statistical
When retrieve one particular image in this large technique. Some of the statistical technique representation
database, and then they cannot access this image unless by of texture feature use technique like, tamura feature, co-
using searching the image, scrolling, browsing and occurance matrices, multi-resolution technique such as
retrieving. One main problem is that it is difficult to Gabor filter and wavelet transformation.
locating or finding a desired image in this large collection
of database. Shape based image retrieval are classified in two
different categories i.e. object segmentation and shape
Image retrieval is the way of searching the image, representation. They also categories the shape
browsing the image and retrieving image from the large representation in two other sub categories i.e. boundary
collection database of digital image. The main objective is based and region based. For this two categories of shape
that to develop a new technique which extract the image representation they use two method one is the Fourier
from this huge database by using the image feature which descriptors and other is moment variant. [4]
is based on colour, shape, texture, etc. [1]
Some of recent research paper or in review paper li
To search image in database use two type of image and ma this two person work on the shape based image
retrieval technique i.e. text based image retrieval and retrieval. They show the shape representation in two
content based image retrieval. different method i.e. geometric moments method (region
based) and Fourier descriptor (boundary based) this two
Text based image retrieval method which used the method is related to the simple linear transformation. After
metadata to retrieve the image. In TBIR use text of image li and ma second research is by babu etal. He compare
which is present in image. For performing this type of performance of both the method (region based
retrieving method for extract the text of image use representation and boundary based representation) which is
OCR(optical character recognition) algorithm is use. After use in li and ma in shape representation. Babu etal combine
IJISRT19MA302 www.ijisrt.com 280
Volume 4, Issue 3, March – 2019 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
this representation. In this paper showed that the combined used to capture the image content i.e text, line, word,
representation outperformed the simple representation. [6] character, etc. In text based image searching OCR gives
better performance.
In content based image retrieval the earliest work is
done by the two person (Ning-san chang and king-san Fu) B. Content based image retrieve
they describe their paper by taking the example of Query-
by-pictorial. color based: Histogram
Histogram is graphical representation of a image. In
The Ning-san chang and king-san Fu, they use low histogram plot graph according to the image pixel value. It
level feature and same system used semantic for content text the image pixel value and draw the graph and this
based image retrieval. They also use the local feature in graph is compare to all other image in database. Create
another persective. The local feature is use to extract histogram for query image find all 256 value of pixel
feature from the image such as hndle the same image, instead of to store image color. By using distance function
matching the image, or image searching that contain the calculate all distant image. Retrieval image 256 scale in
same image/same object or same scene in query image each RCB channel.
with different direction or different viewpoint, they have
different scale, change in illumination, occlusions. [3] shape based: KNN (k nearest neighbors)
KNN is a type of classification algorithm. The
III. METHODOLOGY neighbor are classified by nearest object. Distance of all
image is calculated from different formula like manhattan,
A. Text based (Algorithm: OCR (optical character eculidean distance. Number of output image is depend on
recognition) ‘K’ value. If suppose value of K is equal to 2 then their is
OCR algorithm is use to scan the text of image and two image in output.
this text can be use to modified the data. OCR extract the
text from the image. It capture the image text a via pattern texture based: Gabor-filter
either word by word or letter, line, etc. OCR algorithm is
IV. ARCHITECTURE
Fig 1:- Architecture of content-based image retrieval
User query: There is user requirement viz, size, type, Image database: It include image detail such as size,
date, time of capture, texture, colour. type, date and time.
Query processing: The image details can be either Feature database: It contain the information of the
manually supplied or generated automatically using extracted feature of input image.
suitable image processing technique. The input query is Matching: These processed quires are then matched
then processed. with the stored data based and the query images are
Feature extraction: It extract the feature from input retrieved. The resulting images is based on the queries
image such as text, colour, shape. entered by the user.
IJISRT19MA302 www.ijisrt.com 281
Volume 4, Issue 3, March – 2019 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
V. WORKING
User can get input in the form of user query, when
user have to search image or query goes input to the query
processing. In this block the user query is processed using
suitable image processing technique. After image
processing, query optimization is use to optimize the
dataset. Then the feature is extract in input image i.e. text,
colour, shape, texture, etc and this feature is check or
compare to the image database and feature database. After
performing all this operation match the image database
feature and feature database and at last after performing
this image retrieval technique we have only similar data,
shoe this data as result.
Fig 4:- OCR Perform for selected image
VI. RESULT
Here is window after clicking on search button. there
is one panel that show the browse image and here is one
button "perform text extraction" after clicking on this
button image text is extracted and shown in text area below
the button.
Fig 2:- Front View
As shown above fig. here it is interface of our project
there are 3 buttons reset button is used to remove all text or Fig 5:- browse image for retrieving color image
image from text field and panel. Browse button is used to
browse image and path of images is shown in text field. In this browse the image for performing color base
Search button is used to search the browse image. image retrieving. Click on browse button after clicking
dialog box is open in this path of image is given, where the
image is store then choose the color image which color
image you have to extract. Then click on open button. In
right panel image preview is shown, which is use to help
select required image.
Fig 3:- Browse image for performing OCR
Here the image is browse when we click on browse
button dialog box is open and here is path of imaged from
that file we select one image. In right panel image preview
is shown, which is use to help select required image. Fig 6:- Search query image
IJISRT19MA302 www.ijisrt.com 282
Volume 4, Issue 3, March – 2019 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
After browse the image for color-based image REFERENCES
retrieval, this window is open. Here two button is available
i.e. “perform text extraction” and “search”. By clicking on [1]. Wengang Zhou, Houqiang Li, and Qi Tian Fellow,
“perform text extraction” button text is extract if text is “Recent Advance in Content-based Image Retrieval: A
present in input image. And after clicking the “search” Literature Survey” IEEE,2017
button the similar image (similar color image) is found. [2]. Mohammed Alkhawlani, “Text-based, Content-based,
and Semantic-based Image Retrievals: A Survey
Dept.of Information Systems”, Faculty of Computers
and Information, Mansoura University Mansoura,
Egypt, 01 January 2015
[3]. T.Karthikeyan,P.Manikandaprabhu,
S.Nithya,Associate, “A Survey on Text and Content
Based Image Retrieval System for Image Mining” ,
Department of Computer Science ,PSG College Of
Arts and Science, Coimbatore, Tamilnadu, India.3
March 2014
[4]. Amandeep Khokher,Rajneesh Talwar, “Content-based
Image Retrieval: Feature Extraction Techniques and
Applications” RIMT-Maharaja Aggrasen Engineering
College Mandi Gobindgarh,2012
Fig 7:- Output of query image
[5]. Juan Manuel Barrios, Diego D´ıaz-Espinoza, and
Benjamin Bustos, “Text-Based and Content-Based
After click on the search button histogram algorithm
Image Retrieval on Flickr: demo”, Department of
really work and apply histogram algorithm and calculate
Computer Science University of Chile Santiago,
similar image by taking input as query image. List of
Chile,2009
similar image is shown in matlab because as compare to
[6]. Debashish pal, Madhusudhan Mishra, “literature
java matlab give better clarity of image. In final result first
survey on content-based image retrieval (CBIR)
image is query image (input image) and other image is
technology", North Eastern Regional Institute of
similar to that image.
Science and Technology Madhusudhan Mishra, April
2014
Fig 8:- Output of query image for performing shape-based
image retrieval
After performing color-based retrieval next type is
shape based image retrieval steps are same stated above in
color based. After clicking on search button image is
displayed and all this image have same shape as query
image.
VII. CONCLUSIONS
We present system that allow a combined search of
text and content based feature. In text based we provide a
keyword, description or image metadata in side bar. In
CBIR we extract image feature i.e. shape, color, texture.
Using both the technique we improved the performance
speed and reduce searching time. Gives the better result
using matlab and java both interface.
