A Study on Image Retrieval Features and Techniques with Various Combinations

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 04 Issue: 09 | Sep -2017 www.irjet.net p-ISSN: 2395-0072
© 2017, IRJET | Impact Factor value: 5.181 | ISO 9001:2008 Certified Journal | Page 1158
A Study on Image Retrieval Features and Techniques with Various
Combinations
Pooja Tripathi1, Prof. Parikshit Tiwari2
1M.Tech. Scholar, Computer Science Dept. RIT College Rewa, M.P.
2Prof. Computer Science Dept. RIT College Rewa, M.P.
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Nowadays images are the main source of
information available after text; this fact is due to the
availability of inexpensive image registration (e.g.,
photographic cameras and cell phones) and data storage
devices (large volume hard drives), which have give rise tothe
existence of millions of digital images stored in many
databases aroundtheworld. However, storedinformation may
be useless if one cannot access the specific data as per his
interest on. So this paper detailed various methods adopt by
researchers for image retrieval. Here detail reviewofdifferent
image fetching or ranking techniquesisstudy withtheirimage
features.
Key Words: Content Features, data Retrieval, Image
Processing, Image Re-ranking. …
1. INTRODUCTION
In the recent years the progression in computer and sound
advances has prompted the creation of computerized
pictures with picture archives. The extent of picture
accumulations has expanded quickly because of this,
including advanced libraries, medicinal picturesandsoforth.
To handle this quick development it is required to create
picture recovery frameworks which work on a substantial
scale. The essential point is to fabricate a hearty framework
that makes, oversees and inquiry picture databases in an
exact way. CBIR is the method of naturally ordering pictures
by the extraction of their low-level visual highlights, similar
to shape, shading, and surface, and these filed highlights are
exclusively in charge of the recovery of pictures [8]. Along
these lines, one might say that through route, perusing,
query by-case and so forth one can figure the likeness
between the low-level picture substances which can be
utilized for the recovery of important pictures.Picturesarea
portrayal of focuses in a high dimensional component space
and a metric is utilized to gauge the closeness or disparity
between pictures on this space. Subsequently,thosepictures
which are nearer to the query picture are like it and are
recovered. Highlight portrayal and likeness estimation are
exceptionally critical for the recovery execution of a CBIR
framework and for quite a long time scientists have
considered them broadly. An assortmentofprocedureshave
been proposed yet and, after its all said and done itstaysasa
standout amongst the most difficultissuesinthe progressing
CBIR explore, and the primary purpose behind it is the
semantic hole issue that exists betweenthelow-level picture
pixels caught by machines and high level semanticideassaw
by people.
A content based image recovery (CBIR) framework chips
away at the low-level visual highlights of a clientinputquery
picture, which makes it troublesome for the clients to figure
the inquiry and furthermore does not give tasteful recovery
comes about. In the past picture explanation was proposed
as the most ideal framework for CBIR which takes a shot at
the guideline of consequently allocating watchwords to
pictures those assistance picture recovery clients to query
pictures in view of these catchphrases.Picture explanationis
regularly viewed as the issue of picture arrangement where
pictures are spoken to by some low-level highlights and the
mapping between low-level highlights with abnormal state
ideas (class names) is finished by administered learning
calculations. In a CBIR framework learning of compelling
component portrayals and similitude measuresiscritical for
the recovery execution. Semantic hole has been the key test
for this issue. A semantic hole exists between low-level
picture pixels caught by machines and the abnormal state
semantics saw by people.
2. Text-Based and Content Based Image Retrieval
(CBIR)
As picture recovery is done by two procedures initially is
content and other is text if there should be an occurrence of
annotation based recovery pack of words are join with the
picture. Since pack of word contains data like keywords,
heading, arranging codes, and so on [2].Thissortofrecovery
is non-institutionalized as explanation of picture is finished
by a human and according to his dialect commentisfinished.
One more issue is that content depiction is perplexing and
basic so picture portrayal is done that is absolutely base on
individual perspectives. This sort of picture association for
huge dataset is unreasonable where constant observation
pictures are coming one by one [3].
Seeking of picture recovery is finished by utilizing contentof
picture which is nothing aside from the visual
characteristics. As content based recovery assumes an
essential part in the picture mining yet to enhance the
precision of bringing picture from database, content BIR is
finished. Here it is a system which semantically coordinates
the picture characteristics of the pictures in the databasefor
grouping, positioning, and so on [4].

Fig.1 Image retrival by text query.
Fig. 2 Image retrival by visual query.
As human association is required to diminish by utilizing
content or comment which is further enhance by this CBIR,
so this is the principle objective of the work. The computer
must have the capacity to recover pictures from a database
with no human presumption on particular space, (for
example, surface versus non-surface, or indoor versus open
air).
Most significant assignment if there should arise an
occurrence of the content base recovery is challenging the
tenets for the correlation of picture characteristics ormarks
to different pictures in the database. As highlight
examination incorporates pixel esteems. As contrast
between the characteristics is computed for finding the
pertinence as important pictures have less distinction while
unimportant element has high contrast.
3. Related Work
Simardeep Kaur et. al., [1] displayed HSV based color space
picture recovery strategy, in view of the color conveyance of
the pictures. The execution of content based picture
recovery utilizing HSV color space is assessed and after that
RGB and HSV show is thought about. The CBIR utilizing HSV
color space plot exchanges every pixel of picture to a
quantized color and utilizing the quantized color code to
think about the pictures of database. This HSV esteemshasa
high review and exactness of recovery, and is successfully
utilized as a part of content based picture recovery
frameworks.
S. P. James [2] introduced a picture recovery framework
(CBIR), utilizing HSV color characteristics , which can
recover facial pictures from the extricated facial
characteristics . K-Means bunching method is connected to
the pictures are at first grouped into a gathering which has
comparable HSV color content. At that point the picked
assemble is bunched utilizing K-Meansgroupingcalculation.
The trial result is contrasted and Euclidean separation
metric where the bunchingstrategy producesprecisepicture
recovery and better order of pictures.
In [3], the color qualities is expelled from thejointhistogram
in perspective of the mix of the tint and drenching and the
surface component is removed using the GCLM incorporate.
The k-suggests gathering is used to cluster the part of the
photo. The ROC bend is pulled in demand to survey the
execution of the part extraction. The chi-square is used to
find the similarity between the two pictures. The appraisal
happens show the precision of the recuperation in light of
the exactness and audit false positive and negative extent.
The ROC bend is used to consider the profitability of the
color, surface and the blend of both thecolorandthesurface.
Iyad Aldasouqi and Mahmoud Hassan [4], proposed a quick
calculation for distinguishing human faces in color pictures
utilizing HSV color model without trading off the speed of
identification. The calculation is quick and can be utilized as
a part of some ongoing applications.
Vadivel, An et. al., [5], did a point by point investigation of
the properties of the HSV (Hue, Saturation and Intensity
Value) color space, laid accentuation on the visual view of
the shade of a picture pixel with the variety in tint,
immersion and force estimations of the pixel. Utilizing the
aftereffects of this investigation, they decided the relative
significance of tone and force in light of the immersion of a
pixel and connected this idea in histogram era for content-
based picture recovery (CBIR) from extensive databases. In
conventional histograms, every pixel contributes just to one
segment of the histogram. In any case, they proposed a
strategy utilizing delicate choice that adds to two parts of a
histogram for every pixel. Shamik
Sural et. al., [6] inspected the properties of the HSV (Hue,
Saturation and Value) color space with emphasis on the
visual impression of the assortment in Hue, Saturation and
Intensity estimations of a photo pixel. They isolated pixel
incorporates by either picking the HueortheIntensityasthe

dominating property in perspective of the Saturation
estimation of a pixel. The element extraction technique has
been connected for both picture division and in addition
histogram era applications – two unmistakable ways to deal
with content based picture recovery (CBIR). The K-implies
bunching of characteristics joins pixels with comparable
color for division of the picture into objects. The histogram
holds a uniform color progress that empowers them to do a
window-based smoothing amid recovery.
Chun et. al., [8], proposed a content based picture recovery
technique as its surface characteristics , BDIP and BVLC
snapshots of the esteem part picture are received. The color
and surface characteristics areseparatedinmulti-resolution
wavelet space and joined.
Youthful Deok et. al., [9] proposed square contrast of
backwards probabilities (BDIP) and piece variety of nearby
relationship coefficients (BVLC), for content-based picture
recovery and after that introduced a picture recovery
technique in light of the blend of BDIP and BVLC minutes.
Their exhibited recovery technique yields around 12%
better execution in exactness versus review.
Yu-Len Huang et. al. [10] displayed a COMPUTER helped
conclusion (CAD) framework with textural characteristics
for characterizing amiable and harmful bosom tumors on
therapeutic ultrasound frameworks. The proposed CAD
framework used effortless textural characteristics , i.e.,
square distinction of converse probabilities, piece variety of
nearby relationship coefficients and auto-covariancegrid, to
distinguish bosom tumor. The proposed framework
recognizes bosom tumors with a relatively high exactness.
Ying Ai Ju et. al., [11] proposed a face acknowledgment
technique utilizing nearby insights of slopes and
relationships. BDIP (piece distinction of opposite
probabilities) is picked as a neighborhood measurements of
inclinations and two sorts of BVLC (square variety ofnearby
relationship coefficients). The melded characteristics of
BDIP and BVLCs are more strong to variety of light and
outward appearance thus the proposed technique yields
great outcomes. A considerable lot of the surface
characteristics have been producedsofaramidtheprevious
years. According to our writing study over surface
component discovery,oneinferredthatBDIPandBVLCwere
proposed as of late and demonstrated a superior recovery
productivity in different areas including content based
picture recovery.
4. Features of Image Retrieval
Color Histogram: The color histogram delineates color•
characteristics which can't catch color appropriations or
surfaces inside the picture. In this techniquecolor histogram
include is partitioned into Global and neighborhood color
extraction. Utilizing Global Color Histogram,a picturewill be
encoded with its color histogram, and the separation
between two pictures will be dictated by the separation
between their color histograms [6]. Nearby color histogram
gives spatial data. Nearby color histogram likewise givesthe
data identified with the color conveyance of districts. The
initial step is to isolate the picture into section and after that
to get a color histogram for each square at that point picture
will be spoken to by these histograms. When looking at two
pictures, one compute the separation, utilizing their
histograms, between a district in one picture and a locale in
same area in the other picture [6].
Color Moments: Colorminutesarethemeasurablesnapshots
of the likelihood conveyances of hues and have been
effectively utilized as a part of numerous recovery
frameworks, particularly when the picture containsonlythe
query, it implies color minute will work best when picture
has just protest. Three parameters are figured in this
technique: Mean, Variance and Skewness. Color minutes
have been ended up being proficient and compelling in
speaking to color dispersions of pictures and it experience
the ill effects of the issue that they neglect to encode any of
the spatial data encompassing the color inside the picture
[8].
Color Coherence Vector: It is a part histogram which
allotments pixels as indicated by their spatial intelligibility.
Every pixel inside the picture is parceled into two sorts, i.e.,
cognizant or confused rely upon whether it is a piece of a
bigger locale of uniform color.Isolatehistograms wouldthen
be able to be created for both lucid and garbled pixels
consequently incorporating some spatial data in the
component vector and because of itsextra spatial data;ithas
been demonstrated that it gives preferred recovery comes
about over the color histogram [8].
Discrete Wavelet Transform: This strategy is utilized to
deteriorate a flag. In this technique, one decay the flag
utilizing channel banks. The yields of channel banks are
down tested, quantized and encoded by the encoders. The
decoders are utilized to unravel the coded portrayals.Takea
N×M picture at that point channel each line and after that
down specimen to get two N×M/2 pictures at that point
channel every segment and sub test the channel yield to
acquire four sub pictures, the one got by low-pass sifting the
lines and sections is alluded to as the LL picture, the one got
by low – pass separating the lines and high pass sifting the
segments is alluded to as the LH picture, the one got by high
pass sifting the linesandlow-passseparatingthesegmentsis
known as the HL picture and thesubpicturegotbyhigh-pass
sifting the lines and segments is alluded to as the HH picture
and each of the sub-pictures got in this mold would then be
able to be sifted and sub inspected to get four more sub
pictures. This procedure can be proceeded until the coveted
sub band structure is gotten [10, 12].

Edge Feature: As picture is a gathering of power esteems,
and with the sudden change in the estimations of a picture
one imperative element emerges as the Edge as appeared in
figure 4. This element is use for various sort of picture
protest discovery, forexample,expandingona scene,streets,
and so forth [15]. There are numerous calculation has been
created to adequately call attention to every one of the
pictures of the picture or edges which are Sobel, perwitt,
shrewd, and so forth out of these calculations vigilant edge
recognition is outstanding amongst other calculation to
locate every conceivable limit of a pictures.
Fig. 4 Represent Edge feature of an image.
Corner Feature: In order to stabilize the video frames in
case of moving camera it require the difference between the
two frames which are point out by the corner feature in the
image or frame [14]. So by finding the corner position of the
two frames one can detect resize the window in original
view. This feature is also use to find the angles as well as the
distance between the object of the two different frames. As
they represent point in the image so it is use to track the
target object.
Fig 5 Represent the corner feature of an image with green
point.
5. CONCLUSIONS
After analyzing all the techniques of CBIR, paper concluded
that only color or texture or shape feature cannot describe
image so as a suggestion that researcher should find best
method of color feature extraction then best method of
texture feature extraction and then best method of shape
feature extraction. Here various structure of image
arrangement are explained such as clustering, graph,
indexing, etc. It was found that annotation is tough for the
images but increase the relevance accuracy at the time as
compare to content based image retrieval only.
REFERENCES
[1]. S. Kaur and Dr. V. K. Banga, “Content Based Image
Retrieval:Survey andComparison betweenRGBand
HSV”, model AMRITSAR COLLEGE OF ENGG &
TECHNOLOGY, Amritsar, India.
[2]. “Face Image Retrieval with HSV Color Space using
Clustering Techniques” Samuel Peter James
Assistant Professor, Department of Computer
Science and Engineering, Chandy College of
Engineering, Thoothukudi, Tamil nadu, INDIA
[3]. Sridhar, G., “Color and Texture Based Image
Retrieval, (2009).
[4]. I. Aldasouqi and M. Hassan, “Human face detection
system using HSV”, Proc. Of 9th WSEAS Int. Conf. on
Circuits, Systems, Electronics, Control & Signal
Processing (CSECS’10), Atenas, Grecia, (2010).
[5]. A. Vadivel, A. K. Majumdar and S. Sural,
“Perceptually smooth histogram generation from
the HSV color space for content based image
retrieval”, International Conference on Advancesin
Pattern Recognition, (2003).
[6]. Sural, S., G. Qian, and S. Pramanik, "Segmentation
and histogram generationusingtheHSVcolorspace
for image retrieval", Image Processing, 2002,
Proceedings, 2002 International Conferenceon,vol.
2. IEEE, (2002).
[7]. Dahane, G. M., & Vishwakarma, S., “Content Based
Image Retrieval System”, IJEIT, Vol. 1, (2012), Pp.
92-96.
[8]. Chun, Y. D., N. C. Kim, And I. H. Jang, "Content-Based
Image Retrieval Using Multiresolution Color And
Texture Features", Multimedia, IEEE Transactions
On, Vol. 10, No. 6 (2008), Pp. 1073-1084.
[9]. Chun, Y. D., S. Y. Seo, And N. C. Kim, "Image Retrieval
Using BDIP And BVLC Moments", Circuits And

Systems For Video Technology, IEEE Transactions
On, Vol.13, No. 9, (2003), Pp.951-957.
[10]. Huang, Y-L., K-L. Wang, and D-R. Chen,"Diagnosisof
breast tumors with ultrasonictextureanalysisusing
support vector machines", Neural Computing &
Applications, vol. 15, no. 2 (2006), pp. 164-169.
[11]. Ju, Y. A., "Face recognition using local statistics of
gradients and correlations", Proc. European Signal
Processing Conf., (2010).
[12]. Meng Wang, Hao Li, Dacheng Tao, Ke Lu, and
Xindong Wu “Multimodal Graph-Based Reranking
for Web Image Search. IEEE Transaction on image
processing Vol. 21, NO. 11, November 2012.
[13]. G. Bradski and A. Kaehler, Learning OpenCV :
Computer Vision with the OpenCV Library, O'Reilly,
Sebastopol, CA, 2008.
[14]. Herbert Bay, Andreas Ess, Tinne Tuytelaars, Luc
Van Gool, SURF: Speeded Up Robust Features",
Computer Vision and Image Understanding (CVIU),
Vol. 110, No. 3, pp. 346--359, 2008.
[15]. Bindita Chaudhuri, Begüm Demir, Lorenzo
Bruzzone and Subhasis Chaudhuri. Region-Based
Retrieval of Remote Sensing Images Using an
Unsupervised Graph-Theoretic Approach. Ieee
Geoscience And Remote SensingLetters,Vol.13,No.
7, July 2016 987.

A Study on Image Retrieval Features and Techniques with Various Combinations

More Related Content

What's hot (18)

Similar to A Study on Image Retrieval Features and Techniques with Various Combinations (20)

More from IRJET Journal (20)

Recently uploaded (20)

A Study on Image Retrieval Features and Techniques with Various Combinations