Analysis of color image features extraction using texture methods

TELKOMNIKA, Vol.17, No.3, June2019, pp.1220~1225
ISSN: 1693-6930, accredited First Grade by Kemenristekdikti, Decree No: 21/E/KPT/2018
DOI:10.12928/TELKOMNIKA.v17i3.9922 1220
Received May 18, 2018; Revised January 21, 2019; Accepted February 25, 2019
Analysis of color image features extraction using
texture methods
Aws AlQaisi*1
, Mokhled AlTarawneh2
, Ziad A. Alqadi3
, Ahmad A. Sharadqah4
1
Al-Balqa Applied University, Faculty of Engineering Technology,
Communication Engineering Department, Amman, Jordan
2
Mutah University, Faculty of Engineering, Computer Engineering Department, Al-karak, Jordan
3,4
Al-Balqa Applied University, Faculty of Engineering Technology,
Computer Engineering Department, Amman, Jordan
*Corresponding author, e-mail: aws.al-qaisi@bau.edu.jo
Abstract
A digital color images are the most important types of data currently being traded; they are used
in many vital and important applications. Hence, the need for a small data representation of the image is
an important issue. This paper will focus on analyzing different methods used to extract texture features for
a color image. These features can be used as a primary key to identify and recognize the image. The
proposed discrete wave equation DWE method of generating color image key will be presented,
implemented and tested. This method showed that the percentage of reduction in the key size is 85%
compared with other methods.
Keywords: center-symmetric LBP, discrete wave equation DWE, key size, local binary pattern LBP,
reduced LBP
Copyright © 2019 Universitas Ahmad Dahlan. All rights reserved.
1. Introduction
The Digital color images (DCI) are the most important types of data currently being
traded [1]; they are used in many vital applications such as military, civilian and medical
applications. Digital color image is represented by a 3D matrix, the first 2D matrix represents the
red color, the second 2D matrix represents the green color, the third 2D matrix represents the
blue color [2-4]. DCI usually has a high resolution; this means that it has a huge size of data in
terms of processing and manipulating. To reduce the complexity of processing, DCI can be
represented by image histogram. Hence, each color in DCI consists of 256 elements array,
where each element points to the repetition of a gray value (0 to 255) [5-7]. Figure 1 shows a
sample of a color image and its histogram.
Figure 1. Color image and histograms.

TELKOMNIKA ISSN: 1693-6930 
Analysis of color image features extraction using texture methods (Aws AlQaisi)
1221
Color image histograms can be used as keys to retrieve the image. The key size is very
small (6144=256*3*8 bytes) comparing to the color image size [8]. However, it is big in term of
key representation. Hence, another method is needed to find a reduced key length. One of the
simplest methods of reducing the image key (histogram) size is to reshape the 3D color matrix
to 2D gray matrix. Here the key size is reduced from 6144 to 768 (256*3 bytes). Hence,
reshaping the 3D color matrix to 2D gray matrix will result in reducing the key size 8 times.
Figure 2 shows the histogram of the reshaped image shown in Figure 1.
Figure 2. Reshaped color image and histogram
2. Color Image Texture Feature Extraction Methods
Explaining Color is a widely used image feature for image representation, while the use
of color histogram is the most common way for representing color features [9]. A various
methods used for color feature extraction and could be taken in account for key retrieving
element [10]. Looking for a small size key element local binary based methods was tackled.
2.1. Local Binary Pattern Method
Local binary pattern (LBP) method creates a LBP operator for each pixel in the
image [11, 12]. The binary value of this operator is to be converted to a decimal number, which
will form a one repetition of a value from the range (0 to 255). This method works as shown in
Figure 3.
Figure 3. LBP operator calculations
LBP method creates a new histogram (key) with 256 elements. As a result, it does not
suit image retrieval or recognition [13, 14]. However, it could suite other image applications
because the used key for image retrieval is still having a big size [15]. LBP methods can be
implemented by reshaping the 3D color image to 2D gray image, then we have to calculate LBP
operator for each pixel in the 2D matrix as shown in Figure 3. The calculated LBP key for the
image shown in Figure 1 is demonstrated in Figure 4.

 ISSN: 1693-6930
TELKOMNIKA Vol. 17, No. 3, June 2019: 1220-1225
1222
Figure 4. LBP key for the image in Figure 1
2.2. Center-Symmetric Local Binary Pattern Method
Center-Symmetric Local Binary Pattern method (CSLBP) of color image features
extraction (key generation) is a modified version of the LBP method [11-14, 16] . This method is
used to create a key for a color image by reshaping a 3D color matrix into 2D matrix [17].
Then a CSLBP operator must be calculated for each pixel in the 2D matrix as shown
in Figure 5.
Figure 5. CSLBP operator calculations
This method reduced the key length to 16 elements and the size to 128 bytes. This
means that the CSLBP method makes an improvement to generate keys. Figure 6 shows the
key (histogram) for the image shown in Figure 1
Figure 6. CSLBP key for the image in Figure 1
2.3. Reduced Local Binary Pattern Method
Reduced local binary pattern method (RLBP) acts as CSLBP method. However, the key
will be reduced to 8 elements [18-20]. Figure 7 shows how to calculate RLBP operator for each
pixel in the 2D matrix. Figure 8 shows the generated key for the image shown in Figure 1.

1223
Figure 7. RLBP operator calculations
Figure 8. RLBP key for the image in Figure 1
2.4. The Proposed Method
Discrete wave equation DWE was used to reduce the key length in the smallest
representation. DWE method is based on using wave equation to generate a key for a color
image [21-23]. The wave equation takes the following form as shown in (1):
∂2u(x,t)
∂t2 = c2 ∂2u(x,t)
∂x2 (1)
where 𝑐 = √𝐸 𝜌⁄ is the velocity.
Usually a discrete equation is considered in the form of the equation with finite
difference of second order as shown in (2):
∂2u 𝑛(t)
∂t2 =
c2
ℎ2 (u 𝑛+1(t) − 2u 𝑛(t) + u 𝑛−1(t)) (2)
if we take c=1 and h=1, the (2) can be written as
∂2u 𝑛(t)
∂t2 = (u 𝑛+1(t) − 2u 𝑛(t) + u 𝑛−1(t)) (3)
in (3) can be solved by applying convolution between the array x= [1 -2 1] and the voice signal
(Laplace operator).This method can be implemented by applying the following steps:
1. Get the original color image data matrix.
2. Reshape the 3D data matrix which represents the color image to one row array.
3. Apply convolution between Laplace operator and the row array.
4. Check each value in the convolution results:
a. If the value is greater than zero add 1 to local minimum count.
b. If the value is equal to zero add 1 to stable count.
c. If the value is less than zero add 1 to local maximum count.
5. Save the 3 counts as a features array (key) for the certain voice signal.Using this method,
the color image key reduced to 3 elements (size of 24 bytes).

 ISSN: 1693-6930
TELKOMNIKA Vol. 17, No. 3, June 2019: 1220-1225
1224
3. Implementation
The proposed DWE was implemented in Matlab environment using different color
images type with different size, for each color image we create features key, the results of
implementation are shown in Table 1. From Table 1, it can be concluded the following facts:
 DWE method can be used to generate a key for any color image with any type and size.
 The key is a unique, thus we can use it as a primary key to retrieve or recognize the image.
 The key values are very sensitive to any changes in the image, any change in the image,
even it is a very small change will cause a change in the key.
 The generated key is small in size and contains the values with total size equal 24 bytes.
The other conventional methods of color image key creation were also implemented
and each method gave a unique key. The key generation time for each method was
programmable calculated and the results are shown in Table 2. From Table 2, it can be noted
that the DWE has good timing characteristics. The generated key can be used to identify or
recognize a certain color image. Hence, the key size plays an important role. Table 3 shows a
comparison between the key (size and length) generated by different methods.
From Table 3, it can be concluded that the generated key using DWE method has the
minimum number of size and length compared with other methods. Hence, using the proposed
method will reduce the efforts of dealing with a recognition tool such as artificial neural network
(ANN) [24, 25]. This will result in reducing ANN architecture, reducing memory space size and
reducing ANN training time.
Table 1. Results of DWE
Implementation
Image Size(pixels) Feature array
1 76800 34360 6492 35948
2 270948 110441 54446 106061
3 151875 62371 26119 63385
4 49152 23633 1324 24195
5 1125600 547680 34365 543555
6 540000 254064 35688 250248
7 3396069 1569304 296553 1530212
8 2359296 1050979 288046 1020271
9 928800 456157 35177 437466
10 432000 203944 23608 204448
Table 2. Key Generation Time (Seconds) For
Various Methods
Image LBP CSLBP RLBP DWE
1 0.009000 0.005000 0.004000 0.019000
2 0.031000 0.017000 0.013000 0.032000
3 0.016000 0.010000 0.007000 0.020000
4 0.007000 0.004000 0.003000 0.012000
5 0.136000 0.077000 0.064000 0.096000
6 0.062000 0.038000 0.029000 0.051000
7 0.393000 0.262000 0.191000 0.270000
8 0.280000 0.169000 0.134000 0.194000
9 0.107000 0.062000 0.049000 0.081000
10 0.050000 0.029000 0.024000 0.042000
Table 3. Key Length (Elements) and Size (Byte) for Various Methods
Method Key length Key size
LBP 256 2048
CSLBP 16 128
RLBP 8 64
DWE 3 24
4. Conclusion
Different methods of color image key generation were implemented and tested for
various color images with various types and sizes. The obtained results show that by using the
LBP method the key size is 256 elements, while the key size is reduced to 16 elements by using
CSLBP method. By using the RLBP method, the key size is reduced to 8 elements.
The interesting finding is that the proposed DWE method achieves the minimum key size of 3
elements. It was shown that the proposed DWE method is the best among conventional
methods in terms of the generated key. Hence, the complexity of the image recognition process
will be minimized using the proposed method.
References
[1] M Kriss, M Kriss. Ed. Handbook of Digital Imaging. United Kingdom: John Wiley & Sons, Inc,
2015. 1824.
[2] Z AlQadi and HM Elsayyed. Window Averaging Method to Create a Feature Victor for RGB Color
Image. International Journal of Computer Science and Mobile Computing. 2017; 6(2): pp. 60-66.

1225
[3] M AlZudool, S Khawatreh, Z Alqadi. Efficient Methods used to Extract Color Image Features.
International Journal of Computer Science and Mobile Computing. 2017; 16(12): 7-14.
[4] J Nader, Z Alqadi, B Zahran. Analysis of Color Image Filtering Methods. International Journal of
Computer Applications. 2017; 174(8): 12-17.
[5] Z Alqadi, A Moustafa, M Alduari, R abuZneit. True Color Image Enhancement Using Morphological
Operations. International Review on Computers & Software, vol. 5, no. 4, pp. 557-562, 2009.
[6] M. Al-Dwairi, Z. Alqadi, A. AbuJazar, and R. A. Zneit, "Optimized True-Color Image Processing,"
World Applied Sciences Journal vol. 8, no. 10, pp. 1175-1182, 2010.
[7] A. Sakila and S. Vijayarani, "Image Enhancement using Morphological Operations," International
Journal of Scientific Research in Science, Engineering and Technology, vol. 3, no. 2, pp. 685-698,
2017.
[8] R. AbuZneit, Z. AlQadi, and M. AbuZalata, "A Methodology to Create a Fingerprint for RGB Color
Image," International Journal of Computer Science and Mobile Computing, vol. 16, no. 1, pp. 205-212,
2017.
[9] G. Tripathi, "Review on color and texture feature extraction techniques," International Journal of
Enhanced Research in Management & Computer Applications, vol. 3, no. 5, pp. 77-81, 2014.
[10] D. Srivastava, R. Wadhvani, and M. Gyanchandani, "A Review: Color Feature Extraction Methods for
Content Based Image Retrieval," International Journal of Computational Engineering & Management,
vol. 18, no. 3, pp. 9-13, 2015.
[11] R. Gupta, H. Patil, and A. Mittal, "Robust order-based methods for feature description," in IEEE
conference on computer vision and pattern recognition (CVPR), USA, 2010.
[12] X. Hong, G. Zhao, M. Pietikäinen, and X. Chen, "Combining LBP difference and feature correlation for
texture description," IEEE TRANSACTIONS ON IMAGE PROCESSING, vol. 23, no. 6, pp. 2557–
2668, 2014.
[13] J. Meng, Y. Gao, X. Wang, T. Lin, and J. Zhang, "Face Recognition based on Local Binary Patterns
with Threshold " presented at the IEEE International Conference on Granular Computing, USA, 2010.
[14] M. Pietikäinen, A. Hadid, G. Zhao, and T. Ahonen, "Local Binary Patterns for Still Images," in
Computer Vision Using Local Binary Patterns, vol. 40: Springer-Verlag London Limited, 2011, pp. 13-
47.
[15] K. Arai, Y. Herdiyeni, and H. Okumura, "Comparison of 2D and 3D Local Binary Pattern in Lung
Cancer Diagnosis," International Journal of Advanced Computer Science and Applications, vol. 3, no.
4, pp. 89-95, 2012.
[16] H. Rami, M. Hamri, and L. Masmoudi, "Objects Tracking in Images Sequence Using Center-
Symmetric Local Binary Pattern (CS-LBP)," International Journal of Computer Applications
Technology and Research, vol. 2, no. 5, pp. 504-508, 2013.
[17] R. Davarzani, S. Mozaffari, and K. Yaghmaie, "Image authentication using LBP-based perceptual
image hashing," Journal of AI and Data Mining, vol. 3, no. 1, pp. 21-30, 2015.
[18] A. Porebski, N. Vandenbroucke, and L. Macaire, "Haralick feature extraction from LBP images for
color texture classification," presented at the First Workshops on Image Processing Theory, Tools and
Applications, Tunisia, 2008.
[19] Y. Zhao, W. Jia, and H. Min, "Completed robust local binary pattern for texture classification,"
Neurocomputing, vol. 106, pp. 68-76, 2013.
[20] J. Kwak, S. Xu, and B. Wood, "Efficient Data Mining for Local Binary Pattern in Texture Image
Analysis," Expert Syst Appl, vol. 42, no. 9, pp. 1-29, 2015.
[21] A. Fichtner and S. Hanasoge, "Discrete wave equation upscaling," Geophysical Journal International,
vol. 209, no. 1, pp. 353–357, 2017.
[22] V. E. Tarasov, "Discrete Wave Equation with Infinite Differences," An International Journal of Applied
Mathematics & Information Sciences Letters, vol. 5, no. 2, pp. 41-44, 2017.
[23] E. Vega, A. Orozco, L. Villalba, and J. Castro, "Digital Images Authentication Technique Based on
DWT, DCT and Local Binary Patterns," Sensors, vol. 18, no. 10, pp. 1-18, 2018.
[24] K. Matrouk, H. Alashaary, A. Alhasanat, and Z. Alqadi, "Investigation and Analysis of ANN
Parameters," European Journal of Scientific Research, vol. 121, no. 2, pp. 217-225 2014.
[25] O. Abioduna, A. Jantana, A. Omolarac, K. Dadad, N. Mohamed, and H. Arshad., "State-of-the-art in
artificial neural network applications: A survey," Heliyon, vol. 4, no. 11, pp. 217-225, 2018.

Analysis of color image features extraction using texture methods

More Related Content

What's hot (20)

Similar to Analysis of color image features extraction using texture methods (20)

More from TELKOMNIKA JOURNAL (20)

Recently uploaded (20)

Analysis of color image features extraction using texture methods