LQEBP

Journal of Theoretical and Applied Information Technology
31st January 2018. Vol.96. No 2

© 2005 – ongoing JATIT & LLS
ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195
LOCAL QUANTIZED EDGE BINARY PATTERNS FOR

COLOUR TEXTURE IMAGE RETRIEVAL
G. SUCHARITHA1 RANJAN K. SENAPATI2
1
Department of Electronics and Communication, KL University, India
2
Department of Electronics and Communication,KL University, India
1
sucharithasu@gmail.com, 2ranjan.senapati@kluniversity.in
ABSTRACT
A new image retrieval technique has been proposed in this paper namely local quantized edge binary
patterns using HSV color space. To make use of color, intensity and brightness of image HSV color space
is used. Local quantized patterns are used to define the local information in all possible directions
i.e 0 , 45 , 90 and135 . Local edge binary patterns extract the maximum edge information from the
quantized patterns in all directions in an image. Performance is compared with local binary pattern(LBP),
Centre symmetric local binary pattern(CS_LBP) with color feature, local edge patterns for segmentation
(LEPSEG) with color feature and wavelet transform with color feature and another existing method by
conducting three experiments on standard databases Corel-10k, MIT-Vistex and Color Brodatz texture.
After results investigation, a significant improvement in terms of their evaluation measures as compared
with existing methods on respective databases.
Keywords: Content Based Image Retrieval, Local Binary Patterns, HSV Color Space, Local Quantized
Pattern, Edge Binary Patterns, Histograms, Feature Vector.
1. INTRODUCTION
Now a days it is observed that, the multimedia pixel intensity values. Ahmaddinet.al[6] used the
libraries expanded radically in the internet due to Wavelet transmission (WT) for texture
high quality cameras, scanners etc. Handling of classification. Moghahaddam et.al [7] introduces a
these massive digital libraries are extremely new algorithm called Wavelet correlogram (WC).
annoying rather impractical task. On the other hand, Kokare et.al [8,9] proposed rotated and complex
there exist two major difficulties, especially when rotated wavelet filters to be used for CBIR.
the size of image collections is vast. To defeat these Subramanyamet.al[10] proposed correlogram
difficulties, content based image retrieval (CBIR) algorithm for image retrieval using wavelet and
was proposed. In CBIR, the feature extraction is a rotated wavelet. The idea of color histogram was
important step and the success of a CBIR system proposed by Swain et.al[11] for image matching
depends typically on the method of feature and distance measure by histograms.
extraction from raw images. Hence, there is a great Strickeret.al[12] introduced two new methods for
need of some expert technique viz content based color indexing, in that first gives the complete color
image retrieval (CBIR). distribution and second holds only major features.
Compositions of color and texture features are used
1.1 Background: The CBIR utilizes the low level as a dominant descriptor for image retrieval.
features of an image such as color, shape, texture, Wavelet and Gabor wavelet transform [13]
spatial layouts and faces etc, in order to represent combined with color histogram for color and
and index the image. However, complete and texture based image retrieval. Further, Lin.et.al [14]
extensive literature survey in [1-4]. Among all these introduced color co-occurrence matrix (CCM),
local features texture classification is an active color histogram for K-mean and (CHKM) and
research topic in the fields of medical, object based difference between pixels of scan patterns
image coding, image retrieval and many more. (DBPSP). Instead of RGB, HSV color space is
Various algorithms have been proposed for texture used for color representation and this method is
analysis.Chellepaet.al[5] used the Gaussian Markov applied for image retrieval [15]. S.Jeonget.al[16]
Random fields (GMRF) to model texture pattern used Gaussian mixer vector quantization for better
based on statistical relationship between adjacent quantization of color histogram for image retrieval.
291
In addition to the previous ones, the statistical local In section1 literature survey, which includes
image feature extraction proved to be very efficient motivation and main contribution of proposed
in recent years. Ojala et.al [17,18] introduced local work, section2 involves the introduction of color
binary pattern(LBP) to extract the local information space and various texture descriptors LBP and
of each pixel using eight neighbouring pixels. LBP LMEBP. In section3 proposed method, similarity
was modified for rotational invariant texture measurements and framework of proposed method.
classification [19]. LBP used for many signal and Section 4 gives experimental results and
image processing applications eg, Face discussions on two different databases. As a final
recognition[20], texture classification[21,22], object point, the conclusion is achieved in section5.
tracking[23]etc. Heikkilaet.al[24], proposed the
modified LBP as center symmetric local binary 2.LOCAL PATTERNS
pattern (CS_LBP) which computes the difference in
only four directions. Some other extensions in LBP 2.1 Color Space: In nature, there are 3 types of
were introduced for better feature extraction like images, Binary images, Grayscale images and
completed LBP[25], Dominant LBP [26], local Color images. Binary images contain only two
edge patterns for segmentation and rotation intensity levels for black and white pixels.
invariant image retrieval (LEPSEG& LEPINV) Grayscale images have a range of intensities in one
[27]. Tan.et.al[28] proposed another approach for specific band. The last color images have multiple
local feature extraction for face recognition was bands and each band includes a range of intensity.
local ternary pattern (LTP). That computes the In general color images use RGB color bands called
center pixel and its eight neighbouring pixels with a red, green and blue. Hence it is called RGB color
threshold interval, and assigns a ternary pattern. space. These three bands contain information about
Zhang et.al[29] proposed a higher order local red, green and blue of an image. The other color
binary pattern called local derivative pattern (LDP) space called HSV stands hue, saturation, and value.
for face recognition. Hussainet.al[30,31] proposed
the local quantized patterns for visual recognition. Hue is directly related to color and hue is defined as
It inherits the flexibility and robustness compare to an angle. Saturation represents the lightness and
the traditional local pattern features. It can handle brightness of color segment, and value shows the
significantly larger patterns than previous local intensity of color component. Hue gives an angle
patterns. Subramanyamet.al[32] introduced a new information from 00 to 3600, and each degree
approach based on edge distribution called local occupies different colours. The brightness of an
maximum edge binary pattern (LMEBP). In [33-36] image represented by saturation ranges from 0 to 1,
local tetra patterns (LTrP), directional binary as theintensity of color increases it goes from low
wavelet patterns (DBWP), local mesh patterns and to high. Value also ranges from 0 to 1. Many
local ternary co-occurrence patterns discussed for researches proved that individual RGB components
CBIR and biomedical image are not usually recommended and HSV color model
retrieval.Muralaet.al[37] proposed a new texture is more appropriate than RGB model. In the
feature descriptor called local extrema patterns, proposed method RGB image converted into HSV
where it collects the texture information in all color space.
possible directions of a pixel. Vermaet.al[38]
2.2. Local Binary Pattern (LBP)
proposed color texture feature descriptor called
local extrema co-occurrence pattern. The concept of LBP introduced by Ojala.et.al[17]
for feature classification from neighbourhood
The concepts of LQP and LMEBP have motivated
pixels. It succeeded in many research areas like
us to propose the local quantized edge binary
face recognition, image retrieval etc, due to its
patterns(LQEBP) for image retrieval application.
performance and speed. For a center pixel in 3x3
The main contributions of this algorithm are
pattern, LBP value is calculated by comparing its
summarized as follows. a)The LQEBP features are
gray scale value with its neighbours as
extracted for database images as well as for query
image in contrast of LBP. b) The combination of ∑ )
,
LQEBP and color features is proposed. c) The
performance of the method is experienced for
(1)
image retrieval on various color texture databases.
1 0
The organization of the paper is as follows. (2)
0 0
292
Where P stands for no.of neighbours and R stands deeper quantization with domain-adaptive vector
radius of neighbourhood. quantization.
2.3. Local Quantized Pattern (LQP) It collects all possible geometric features
0 , 45 , 90 135 i.e horizontal(H),
LQP proposed by Hussainet.al[30] for visual vertical(V), diagonal(D) and anti-diagonal(A) strips
recognition. It is a generalized form of local of pixels. Figure2 illustrates geometric structure for
patterns that uses large local neighbourhoods and the LQP operator.
Fig1: Calculation of LBP
Fig 2: Structure of Local Quantized Pattern LQP
2.4. Local Maximum Edge Binary Patterns possible directions using quantized patterns. The
(LMEBP) first maximum edge is obtained by the magnitude
of local difference between the center pixel and its
Subrahmanyamet.al[32] proposed this method for a sixteen neighbors in 0 , 45 , 90 135
given image.The first maximum edge is obtained by for radius equal to 1 and 2 as shown in Figure 4.
the magnitude of local difference between the
center pixel and it’s all neighbors. After calculation
of differences all the values are arranged in
descending order using only magnitudes as shown ∀ 1,2, … … . 8; 1
in Figure 3. Similarly, the remaining seven patterns (3)
are calculated to attain first maximum edge. The
total eight maximum edges are evaluated using nine
binary values.
∀ 16; 2
(4)
3. PROPOSED FEATURE DESCRIPTOR sort the values in descending order
3.1. Local Quantized Edge Binary Pattern | 1 |, | 2 |, …
(LQEBP)
| 8 |, | 1 , 3 ,| 5 , …
The concepts of LBP,LQP and LMEBP motivate us
to propose local quantized edge binary patterns for
15 )) (5)
image retrieval system. The LQEBP extracts the
extreme edge information from an image in all
293
If the value is greater than or equal to 0, assign '1'

otherwise '0'. 1 , 2 , 3 , 4 , , ..
5 ,…… 8 (8)
(6) Now, the entire image converted to LQEBP image

having the values from 0 to 511. After calculation
1 0 the given image represented by building a
(7)
0 histogram supported by Eq.(9).
LQEBP defined as ∑ ∑ , , ; ∈
0,511 (9)
Fig 3: LMEBP calculation
3.2. Analysis of proposed method binary pattern (0 0 1 1 1 01 1 1 0 1 0 1 1 0 0) based

on their signs, similar procedure take on for all
The LQEBP eight patters are highlighted with eight neighbours. Finally, the center to eighth
reference pixels as shown in Figure 4. For the first LQEBP patterns (total in nine) performed as shown
pattern Figure 4(a), the local differences between in Table 1. Total sixteen edges are calculated for a
center pixel to its sixteen neighbours calculated 3x3 pattern. Figure 5represent the flowchart of the
using Eq.3&4. The difference pixel values "13,16,- proposed method and algorithm for the same is
21,-1,-5,9,-17,-9,10,2,7,8,3,-1,4,-3" are arranged in given below.
descending order with respect to magnitudes "-21,-
17,16,13,10,-9,9,8,7,-5,4,-3,3,2,-1,-1" as shown in
Figure 4(j). These sixteen edges are converted into
294
Fig 4: Example Of Obtaining LQEBP For The Matrix 3x3
Table 1: Calculation Of LQEBP For A Given 3x3 Matrix
295
3.3 Similarity measure and Query matching: 3.4 Proposed system framework for image
Feature extraction has to be calculated for all retrieval.
images including the query image, and a feature
vector database has been constructed for the full Figure 5 illustrates the proposed image retrieval
database images. After completing the feature system framework is specified below:
extraction process, similarity has to be performed
for query image. In this paper three types of 1. Load the image and convert it into HSV
similarity distance measures are used as discussed color space.
below: 2. Construct the hue, saturation histograms
separately.
3. Collect the HVDA5 structure for a given
1 : , ∑ | | (10)
center pixel in the value space.
4. Compute the local differences in 0°, 45°,
: , 90°, and 135° directions.
∑ | |(11) 5. Calculate the 16-bit LQEBP pattern for
center pixel and its eight neighbours.
6. Construct the histogram for all LQEBP
: , ∑ | patterns.
| 7. Construct the feature vector by
concatenate all the histograms, i.e Hue,
Saturation and Value.
(12) 8. Compare the query image with images in
the database using Eq. (10).
Where q is the query image, b is the database image 9. Retrieve the images based on the best
matches.
Fig 5: Block Diagram Of Proposed Method.
4.EXPERIMENTAL RESULTS: Average retrieval precision are calculated using

Eq.(13)-Eq.(17).
The proposed method is evaluated on three standard The precision defined for a query image Iq is
databases followed by brief description about Precision ,
experimental conditions. Existing methods ∑| |
, | , |
connected to color and texture have been compared
with the proposed technique, and renowned (13)
measures precision, recall calculated for all
databases for all methods [36]. Precision and recall Where 'n' is the number of top image matches, f(x)
are strong measures for image retrieval but they are is the category of 'x', , returns the rank
individual. To relate these two measures, F-measure of image Ik for the query image Iq, from the
is defined. In every experiment, each image as a database |DB|.
query image and retrieval performance is analyzed. 1
, (14)
Precision, recall, Average retrieval rate, and 0
296
Similarly, recall defined as

Method Feature vector length
Recall ,
∑| |
, | , | (15) Wavelet+ Color 24+192=216
CS-LBP+ Color 16+24=40
LEPINV+ Color 72+24=96
Where, is the number of relevant images in the
LEPSEG+ Color 512+24=536
database. The average retrieval rate(ARR) and
LMEBP 8X511=4088
average retrieval precision(ARP) are calculated
PM(H18-S10-V) 18+10+16x511=8204
using following equations.
PM(H18-S20-V) 18+20+16x511=8214
∑| | , | (16)
| | PM(H36-S10-V) 36+10+16x511=8222
and PM(H36-S20-V) 36+20+16x511=8232
∑| |
, (17)
| |
To measure the capability of the proposed method,

Where, |DB| total number of images in the
experiments have been done on two databases,
database.
where the first one is Corel-10k and second one
F-measure is defined as a relation between MIT-Color database. The table 2 is giving the
precision and recall[38]. It is defined as shown in length of feature vectors for a query image
Eq(18). including the proposed method and state of art
techniques.
. .
_ (18)
. . 4.1 Database1: The Corel-10k[39] database is
larger and adaptable than other Corel databases. It
Table 2: Feature vector lengths for a given comprises of 10000 images of 100 categories,
query image for several methods. where each category has 100 images. It includes
images of animals, e.g fox, tiger, deer etc., human,
natural scenes, ships, food, buses etc., army, ocean,
cats, airplanes etc. Size of images in the database is
85x128. Precision, recall, , and F-
measure are calculated according to the Eq(13)-
(18).A significant improvement has been observed
as compared to the state-of-art techniques as shown
in Fig(6). The image retrieval precision and recall
are shown in Figs 6(a) and 6(b), similarly average
retrieval precision, average retrieval recall and F-
measure are shown in Figs 6(c)-6(e).
Wavelet+Color CS_LBP+Color LEPINV+Color LEPSEG+Color LECoP+Color PM
100
80
% Precision
60
40
20
0
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100
No.of Image Catgory
(a)
297
Wavelet+Color CS_LBP+Color LEPINV+Color

LEPSEG+Color LECoP+Color PM
100
80
% Recall
60
40
20
0
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100
No.of Image Category
(b)

70
60
50
ARP
40
30
20
10
10 20 30 40 50 60 70 80 90 100
No.of images Retrieved
( c)

30
25
20
ARR
15
10
5
0
10 20 30 40 50 60 70 80 90 100
(d)
298
0.3
0.25
F-Measure 0.2
0.15
0.1
0.05
0
10 20 30 40 50 60 70 80 90 100
No.of Top matches
(e)
Figure6: Corel-10k Database:(A) Precision And Image Category Number (B)Recall And Image Category (C) ARP And
Images Retrieved (D) ARR And Images Retrieved (E)F-Measure Vs Top Matched Images.
4.2 Database 2: MIT- Vistex database is used to 640x640 size. Each image is divided into 25 sub
evaluate our proposed method. Figure (7) shows an images of size 128x128 like MIT-Vistex database.
example where apiece image from each category of So, the total number of images in the database is
MIT-Vistex database [40].It comprises of 40 2800. It is comparatively bigger than MIT-Vistex
different colour texture images and each in size of database. Sample images from the database is
512x512. For image retrieval, these images are shown in Fig.(9).The results are show in Fig (10).
divided into 16 blocks where each block size is In Fig.10(a), the graph of precision with number of
128x128, therefore 640 (40x16) image database has images retrieved for a group of 25, and in Fig.10(b)
been created. From Fig.8(a)&8(b) the retrieval graph of recall with number of images retrieved
performance of the proposed method compared from the whole database is shown. The results
with the state- of- art approaches in terms of ARP, determined in graphs openly indicate that the
ARR and F-measure in Fig.8(c). Table 3 shows the proposed method is better than the existing local
results of proposed method and previous methods patterns.
on all colour and texture databases. It demonstrated
that the proposed method showed a significant
retrieval.
4.3 Database 3: Color Brodatz texture database[41]

is used. It has 112 color and texture images of
Figure 7: Sample images from MIT-Vistex Database.
299
Wavelet+Color CS_LBP+Color Wavelet+Color CS_LBP+Color

LEPINV+Color LEPSEG+Color LEPINV+Color LEPSEG+Color
LECoP+Color PM LECoP+Color PM
100 100
90 80
80 60
ARR
ARP
70 40
60 20
50 0
10 20 30 40 50 60 70 80 90 100 10 20 30 40 50 60 70 80 90 100
No.of Images Retrieved No.of Images Retrieved
(a) (b)

0.2
0.15
F-Measure
0.1
0.05
0
10 20 30 40 50 60 70 80 90 100
No.of Top Matches
(c)
Figure 8: MIT-Vistex Database: (A) ARP And Images Retrieved (B) ARR And Images Retrieved (C)F-Measure Vs
Top Matched Images.
Table 3: Results Of Previous Methods And The Proposed Method For The Following Databases.
Corel-10K MIT-Vistex Color-Brodatz
ARP ARR ARP ARR ARP ARR
Wavelet+ 41.5 17.4 87 94 56 70.2
Colorhist
CS_LBP+ 45 19 91 95 58 72
Colorhist
LEPINV+ 45.5 19.4 91 94 59 74
Colorhist
LEPSEG+ 49 21.4 90 95 65 80
Colorhist
LECoP+
52 23.4 99 98 72 83.5
Color
PM 67 27 100 99 76 86
300
Figure 9: Sample Images From Colored-Brodatz Texture Database.

80
Avg.Precision
70
60
50
40
30
20
25 30 35 40 45 50 55 60 65 70
(a)
Wavelet+Colorhist CS_LBP+ Colorhist LEPINV+ Colorhist
LEPSEG+ Colorhist LECoP+Color PM
90
Avg.Recall
80
70
60
50
25 30 35 40 45 50 55 60 65 70
No.of Images Retrieved
(b)
Figure 10: Avg. Precision And Avg. Recall Curves For Color Brodatz Texture Database.
301
5. CONCLUSION & FUTURE WORK: [7] Moghaddam, H. Abrishami, et al. "Wavelet

correlogram: a new approach for image indexing
A new integrated color- texture feature descriptor is and retrieval." Pattern Recognition 38.12 (2005):
proposed for image retrieval. The color image 2506-2518.
converted into HSV color space. Hue and [8] Kokare, Manesh, Prabir K. Biswas, and
Saturation features extracted in various quantization Biswanath N. Chatterji. "Texture image retrieval
levels. The LQEBP applied on value space to using rotated wavelet filters." Pattern recognition
extract the texture features because the vale space is letters 28.10 (2007): 1240-1249.
very near to gray scale image. It extracts the texture [9] Kokare, Manesh, Prabir K. Biswas, and
information in calculating the sixteen maximum Biswanath N. Chatterji. "Texture image retrieval
edges in R=1 and R=2 for each pixel in all possible using new rotated complex wavelet filters." IEEE
Transactions on Systems, Man, and Cybernetics,
directions. Further, the combination of Hue,
Part B (Cybernetics) 35.6 (2005): 1168-1178.
Saturation and LQEBP on value component gives
[10] Subrahmanyam, Murala, R. P. Maheshwari, and
integrated color and texture features. The R. Balasubramanian. "A correlogram algorithm
effectiveness of the proposed method is tested by for image indexing and retrieval using wavelet
conducting experiments for image retrieval on and rotated wavelet filters." International Journal
different image database thereby observed that, of Signal and Imaging Systems Engineering 4.1
significant improvement in terms of their respective (2011): 27-34.
evolution measures. The proposed method feature [11] Swain, Michael J., and Dana H. Ballard.
vector length is very high, if is compressed the "Indexing via color histograms." Active
feature extraction time get reduced. The extension Perception and Robot Vision. Springer Berlin
of this paper may overcome these limitations. Heidelberg, 1992. 261-273.
[12] Stricker, Markus A., and Markus Orengo.
REFERENCES: "Similarity of Color Images." Storage and
Retrieval for Image and Video Databases (SPIE).
[1] Ahonen, Timo, AbdenourHadid, and Vol. 2420. 1995.
MattiPietikainen. "Face description with local [13] Murala, Subrahmanyam, Anil BalajiGonde, and
binary patterns: Application to face RudraPrakashMaheshwari. "Color and texture
recognition." IEEE transactions on pattern features for image indexing and
analysis and machine intelligence 28.12 (2006): retrieval." Advance Computing Conference, 2009.
2037-2041. IACC 2009. IEEE International. IEEE, 2009.
[2] Rui, Yong, Thomas S. Huang, and Shih-Fu [14] Lin, Chuen-Horng, Rong-Tai Chen, and Yung-
Chang. "Image retrieval: Current techniques, Kuan Chan. "A smart content-based image
promising directions, and open issues." Journal of retrieval system based on color and texture
visual communication and image feature." Image and Vision Computing 27.6
representation 10.1 (1999): 39-62. (2009): 658-665.
[3] Smeulders, Arnold WM, et al. "Content-based [15] Vadivel, A., ShamikSural, and Arun K.
image retrieval at the end of the early Majumdar. "An integrated color and intensity co-
years." IEEE Transactions on pattern analysis occurrence matrix." Pattern recognition
and machine intelligence 22.12 (2000): 1349- letters 28.8 (2007): 974-983.
1380. [16] Jeong, Sangoh, Chee Sun Won, and Robert M.
[4] Liu, Ying, et al. "A survey of content-based Gray. "Image retrieval using color histograms
image retrieval with high-level generated by Gauss mixture vector
semantics." Pattern recognition 40.1 (2007): 262- quantization." Computer Vision and Image
282. Understanding 94.1 (2004): 44-66.
[5] Chellappa, Rama, and Shankar Chatterjee. [17] Ojala, Timo, MattiPietikäinen, and David
"Classification of textures using Gaussian Harwood. "A comparative study of texture
Markov random fields." IEEE Transactions on measures with classification based on featured
Acoustics, Speech, and Signal Processing 33.4 distributions." Pattern recognition 29.1 (1996):
(1985): 959-963. 51-59.
[6] Ahmadian, A., and A. Mostafa. "An efficient [18] Ojala, Timo, MattiPietikainen, and TopiMaenpaa.
texture classification algorithm using Gabor "Multiresolution gray-scale and rotation invariant
wavelet." Engineering in Medicine and Biology texture classification with local binary
Society, 2003. Proceedings of the 25th Annual patterns." IEEE Transactions on pattern analysis
International Conference of the IEEE. Vol. 1. and machine intelligence 24.7 (2002): 971-987.
IEEE, 2003.
302
[19] Guo, Zhenhua, Lei Zhang, and David Zhang. "A

completed modeling of local binary pattern
operator for texture classification." IEEE [32] Subrahmanyam, Murala, R. P. Maheshwari, and
Transactions on Image Processing 19.6 (2010): R. Balasubramanian. "Local maximum edge
1657-1663. binary patterns: a new descriptor for image
[20] Ahonen, Timo, AbdenourHadid, and retrieval and object tracking." Signal
MattiPietikäinen. "Face recognition with local Processing 92.6 (2012): 1467-1479.
binary patterns." Computer vision-eccv [33] Murala, Subrahmanyam, R. P. Maheshwari, and
2004 (2004): 469-481. R. Balasubramanian. "Local tetra patterns: a new
[21] Ojala, Timo, MattiPietikäinen, and David feature descriptor for content-based image
Harwood. "A comparative study of texture retrieval." IEEE Transactions on Image
measures with classification based on featured Processing 21.5 (2012): 2874-2886.
distributions." Pattern recognition 29.1 (1996): [34] Murala, Subrahmanyam, R. P. Maheshwari, and
51-59. R. Balasubramanian. "Directional binary wavelet
[22] Ojala, Timo, MattiPietikainen, and TopiMaenpaa. patterns for biomedical image indexing and
"Multiresolution gray-scale and rotation invariant retrieval." Journal of Medical Systems 36.5
texture classification with local binary (2012): 2865-2879.
patterns." IEEE Transactions on pattern analysis [35] Murala, Subrahmanyam, and QM Jonathan Wu.
and machine intelligence 24.7 (2002): 971-987. "Local mesh patterns versus local binary patterns:
[23] Ning, Jifeng, et al. "Robust object tracking using biomedical image indexing and retrieval." IEEE
joint color-texture histogram." International Journal of Biomedical and Health
Journal of Pattern Recognition and Artificial Informatics 18.3 (2014): 929-938.
Intelligence 23.07 (2009): 1245-1263. [36] Murala, Subrahmanyam, and QM Jonathan Wu.
[24] Heikkilä, Marko, MattiPietikäinen, and "Local ternary co-occurrence patterns: a new
CordeliaSchmid. "Description of interest regions feature descriptor for MRI and CT image
with center-symmetric local binary retrieval." Neurocomputing 119 (2013): 399-412.
patterns." ICVGIP. Vol. 6. 2006. [37] Murala, Subrahmanyam, R. P. Maheshwari, and
[25] Guo, Zhenhua, Lei Zhang, and David Zhang. "A R. Balasubramanian. "Directional local extrema
completed modeling of local binary pattern patterns: a new descriptor for content based
operator for texture classification." IEEE image retrieval." International journal of
Transactions on Image Processing 19.6 (2010): multimedia information retrieval 1.3 (2012): 191-
1657-1663. 203.
[26] Liao, Shu, Max WK Law, and Albert CS Chung. [38] Verma, Manisha, Balasubramanian Raman, and
"Dominant local binary patterns for texture SubrahmanyamMurala. "Local extrema co-
classification." IEEE transactions on image occurrence pattern for color and texture image
processing18.5 (2009): 1107-1118. retrieval." Neurocomputing 165 (2015): 255-269.
[27] Yao, Cheng-Hao, and Shu-Yuan Chen. "Retrieval [39] Corel 10k Database, Available online:
of translated, rotated and scaled color 〈http://www.ci.gxnu.edu. in/cbir/〉.
textures." Pattern Recognition 36.4 (2003): 913- [40] MIT Vision and Modeling Group, Cambridge,
929. “Vision texture”, Available online:
[28] Tan, Xiaoyang, and Bill Triggs. "Enhanced local 〈http://vismod.media.mit.edu/pub/〉.
texture feature sets for face recognition under [41] A. Safia, D. He, New Brodatz-Based Image
difficult lighting conditions." Analysis and Databases for Grayscale Color and Multiband
modeling of faces and gestures (2007): 168-182. texture Analysis, ISRN Machine Vision, 2013.
[29] Zhang, Baochang, et al. "Local derivative pattern [Online]. Available:
versus local binary pattern: face recognition with a. <http://multibandtexture.recherche.ushe
high-order local pattern descriptor." IEEE rbrooke.ca/original_brodatz.html>.
transactions on image processing 19.2 (2010):
533-544.
[30] ulHussain, Sibt, and Bill Triggs. "Visual
recognition using local quantized
patterns." Computer Vision–ECCV 2012.
Springer, Berlin, Heidelberg, 2012. 716-729.
[31] Hussain, SibtUl, Thibault Napoléon, and
FrédericJurie. "Face recognition using local
quantized patterns." British machive vision
conference. 2012.
303

LQEBP

Uploaded by

Copyright:

Available Formats

LQEBP

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

LQEBP

Uploaded by

Copyright:

Available Formats

Journal of Theoretical and Applied Information Technology

31st January 2018. Vol.96. No 2

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

LOCAL QUANTIZED EDGE BINARY PATTERNS FOR

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

Fig1: Calculation of LBP

Fig 2: Structure of Local Quantized Pattern LQP

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

If the value is greater than or equal to 0, assign '1'

(6) Now, the entire image converted to LQEBP image

Fig 3: LMEBP calculation

3.2. Analysis of proposed method binary pattern (0 0 1 1 1 01 1 1 0 1 0 1 1 0 0) based

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

Fig 4: Example Of Obtaining LQEBP For The Matrix 3x3

Table 1: Calculation Of LQEBP For A Given 3x3 Matrix

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

Fig 5: Block Diagram Of Proposed Method.

4.EXPERIMENTAL RESULTS: Average retrieval precision are calculated using

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

Similarly, recall defined as

To measure the capability of the proposed method,

Wavelet+Color CS_LBP+Color LEPINV+Color LEPSEG+Color LECoP+Color PM

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

Wavelet+Color CS_LBP+Color LEPINV+Color

Wavelet+Color CS_LBP+Color LEPINV+Color

Wavelet+Color CS_LBP+Color LEPINV+Color

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

4.3 Database 3: Color Brodatz texture database[41]

Figure 7: Sample images from MIT-Vistex Database.

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

Wavelet+Color CS_LBP+Color Wavelet+Color CS_LBP+Color

Wavelet+Color CS_LBP+Color LEPINV+Color

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

Figure 9: Sample Images From Colored-Brodatz Texture Database.

Wavelet+Color CS_LBP+Color LEPINV+Color

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

5. CONCLUSION & FUTURE WORK: [7] Moghaddam, H. Abrishami, et al. "Wavelet

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

[19] Guo, Zhenhua, Lei Zhang, and David Zhang. "A

You might also like