You are on page 1of 12

Preliminary Final Project Assignment

CULTURE DEPENDENT BATIK CLASSIFICATION
WITH ANALYTICAL FUNCTION FOR FEATURE
EXTRACTION


















AYUNINDA DWI NUGROWATI NRP. 7410040044






D4 STUDY PROGRAM OF INFORMATICS ENGINEERING
INFORMATION AND COMPUTER ENGINEERING
ELECTRONIC ENGINEERING POLYTECHNIC INSTITUTE OF
SURABAYA
2013


A. TITLE
Culture-dependent Batik Classification with Analytical Function for Feature
Extraction.

B. INTRODUCTION
Indonesia is a country rich of culture, one of them is batik. Batik is a culture
from Indonesia which had been appointed by UNESCO as “Masterpieces of the Oral
and Intangible Heritage of Humanity”. Batik is an art that has highly value in Indo-
nesia. In the past time, batik is a specific job that can be only done by women. It is
called as a specific job for women because batik production is done using canting as
their tool to draw its pattern in batik and it can only be done by women. Until they
develop “batik cap” that make man can produce it too.
In the current time, batik is highly favored by Indonesian people. Pattern and
color of batik actually is affected by foreign influence. At first, pattern and color ba-
tik is limited. But after foreign culture come to Indonesia, this patterns and colors
increasing. For example a bright color like red is introduced by Tionghoa, and flow-
ers and objects that brought by colonist like horse-drawn carried patterns is intro-
duced by European. Pattern and colors of batik actually has specified meaning and
even some of the pattern only can be used by certain circles. The example is parang
pattern which only can be used by nobles. There is even batiks with specify pattern
that used to culture ceremonies. The variety of pattern and color in batik often affect-
ed by its region.

C. PROBLEM
Batik is an inheritance from ancestor that always keeps from generation to
generation. Batik’s pattern that highly diverse is caused by have specific meaning in
each pattern that inherited. Since their pattern is highly diverse, batik is very hard to
be classified. Each region has their own specify pattern that reflect their identity from
their own region, but it is not rare to have pattern that almost same from one region
to another. It’s not only pattern that make it different, but also from its color that has
specific meaning too. The more variety of batik in Indonesia increases difficulty to
classify these batiks.


D. PROBLEM SCOPE
For the scope of the problem, this application will use batik from east java for
the dataset. This dataset will be captured from image batik from east java and will be
placed on smartphone android that will be used for this research.

E. OBJECTIVE
This research proposes a new system for Batik classification with presenting
culture-dependent Indonesia Batik from several traditional Batik origin places. It
provides an analytical function for feature extraction by involving color and shape
features. 3D-Vector Quantization is applied for color feature extraction. The system
uses Hue-moments to extract shape features. This research presents the system in
mobile application and makes the online classification from camera-capture image.

F. CONTRIBUTION
This research has an advantage for peoples who have interest in batik. This
application makes people who love batik can learn about the origin place of batik
pattern. It uses a mobile phone so that can be easy to carry and use. This application
will provide a place of the origin batik from photos taken by the camera on the mo-
bile phone.


G. DESIGN SYSTEM
This application will divide image into two categories, image database and
image query. Where image database is data image that store in database application
and image query is a new image that will be compared with image database. Figure
of the system which will be made shown below.

Figure 1. Design System

For this research, we use three methods; they are 3D-Color Vector Quantiza-
tion, Hu Moment, and Color Moment. 3D Color Vector Quantization (CVQ) is used
to calculate the weights of the colors in the images of batik. Hu’s Moment is used to
calculate the weight of its shape or motif. And Color moment is used to compare the
Madura
Traditional
Batik
Traditional
Batik
… Batik Design
Mobile Phone
(Image Capture)
Batik Dataset
Batik Image Query
Color Feature
Hu Mo-
ment
Shape Feature

Feature Extraction
Weighting Mechanism
Color Moment
Similarity
Measurement
Batik Origin Place
Jogja
Traditional
Batik

Hu Moment


weight between the images of batik in the database with the query image. To learn
more about all three methods, it will be described below.

a. 3D Color Vector Quantization (CVQ)
For this method, noise removal and 4x4 image partitioning are applied
before extracting color feature. Then, for each block it extract color infor-
mation using the histogram from 3D-Color Vector Quantization of RGB color
space. It use the 64x64x64 quantization size of RGB color space so that it can
be represented with 125 positions in the RGB color space, as shown in Figure
2.

Figure 2. Illustration of 3D-Color Vectore Quantization of RGB color
space

The metadata of color feature MCLb for block b can be described as fol-
lows:

Where:
fci is a color feature of i-th color histogram from 3D-Color Vector Quan-
tization of RGB vector space.
Figure 3 illustrates the mechanism of color feature extraction: image par-
titioning, Color Vector Quantization, and color histogram.



Figure 3. Mechanism of color feature extraction

After extracting the color feature, the metadata of feature is created. This
process is done offline and saved to metadata repository. Because it apply 4x4
image partitioning for each image the metadata of each feature consists of 16
blocks.
They referred to basic concept of Mathematical Model of Meaning
(MMM) for creation of metadata space. The information on data items is given
in the form of matrix. Each data item is provided as fragmentary metadata
which is independently represented one another. The information of each data
item is represented by its features. The n basic data items are given in the form
of an n by m matrix M. for given n basic data items, each data item is charac-
terized by m features. By using this matrix M, the orthogonal space is comput-
ed as the metadata space. Metadata items which are represented in m-
dimensional vectors are mapped into the orthogonal metadata space.
The color metadata (MCL) is shown in below. The attributes consisted of
color metadata from histogram of 3d-Color Vector Quantization of the RGB
vector space[1].




b. Hu’s Moment
The non-orthogonal centralized moments are translation invariant and
can be normalized with respect to changes in scale. However, to enable invari-
ance to rotation they require reformulation. Hu described two different meth-
ods for producing rotation invariant moments. The first used a method called
principal axes, however it was noted that this method can break down when
images do not have unique principal axes. Such images are described as being
rotationally symmetric. The second method Hu described is the method of ab-
solute moment invariants and is discussed here. Hu derived these expressions
from algebraic invariants applied to the moment generating function under a
rotation transformation. They consist of groups of nonlinear centralized mo-
ment expressions. The result is a set of absolute orthogonal (i.e. rotation) mo-
ment invariants, which can be used for scale, position, and rotation invariant
pattern identification. These were used in a simple pattern recognition experi-
ment to successfully identify various typed characters. They are computed
from normalized centralized moments up to order three and are shown below.
Hu invariant moment:



Finally a skew invariant, to help distinguish mirror images, is:


c. Color Moment
This mechanism analyses the distribution of color information to deter-
mine representative features. First, we transform the color space of images into
hybrid color spaces with combining HSL and CIELAB color spaces. The im-
age segmentation is then applied in our image search system using our Pillar-
K-means algorithm.
The system extracts color moments of an image, and calculates the color
distances for the color weight, the texture density for the structure weight and
the shape property for the shape weight. The color moments have been suc-
cessfully used in many retrieval systems and proved to be efficient and effec-
tive in representing color distributions of images. The color moments gives
three kinds of orders, which are the first order (mean µ), the second order (var-
iance σ) and the third order (skewness s).

Where fij is the value of the i-th color component of the image pixel j,
and N is the number of pixels in the image.
To obtain the color weight, the color distances are calculated from the
first order color moment by applying the shape independent clustering [21] in
order to construct and calculate distances of color hierarchy. For measuring the
structure weight, the texture density is calculated from the second order of col-


or moment to be more sensitive to scene the structures of images. The seg-
mented images from the Pillar-Kmeans algorithm are transformed into gray-
scale images to reduce the variance of the second order. To calculate the shape
weight, the shape property is obtained from the third order of color moments.
In this case, the images are converted into binary images in order to sharpen
the skewness. The edges detection is then applied before calculating the third
order of the color moments for shape property. The edges detection is then ap-
plied before calculating the third order of the color moments for shape proper-
ty. The design of the proposed automatic weighting mechanism for our image
search is shown in Figure 6. For the normalization, we set more weighted
consideration for the color weight because the color feature is essential and
dominance to determine the structure and shape weights. In the case of our im-
age search, the color feature is weighted twice rather than the structure and
shape features.

Figure 4. Design of the proposed automatic weighting
mechanism our image search system



H. RELATED WORKS
Batik Indonesia is very loved by the people because of its style and color varie-
ty. Due to a very diverse motif, batik is very difficult to classify. Some researchers
have tried to classify this batik in various ways. One of them is the research from Al-
varez A. Primary, Nanik Suciati, Diana Purwitasari Department of Informatics, Fac-
ulty of Information Technology, Institute of Technology are trying to classify batik
by using Fuzzy C-Means with texture features. In this research, they used the Dis-
crete Wavelet Transform (DWT), Rotated Wavelet Filters (RWF), and Grey Level
Co-occurrence Matrix (GLCM) as the features of images to identify the texture of
batik. Then, the extraction will be grouped into the patterns by using Fuzzy C-Means
clustering (FCM). FCM used directly to label the batik that has more than one pat-
tern gives fairly accurate results [4]. Unfortunately in this research Fuzzy C Means
Clustering is not very suitable for use in clustering data that too spread out like this
batik motif.
Besides that research, there are other researches of classifying batik, one of
them is the research of Dhani Pratikaningtyas, Imam Santoso, Ajub Ajulian Z. on
Batik Classification Method Using Wavelet Transformation Pack. The researchers
use several different types of wavelet to classify batik texture. The types of wavelet
used are Daubechies-2, Daubechies-3, and Coiflet-1. This classification begins with
decomposition process to obtain the wavelet coefficients that will be used to calcu-
late the value of energy and entropy of each image and then combined in the data-
base. The next process is to compare the energy and the entropy between images that
will be classified by the image in the database. The last step is to find the Euclidean
distance to show that image in one of the tests is included in the database Class [5].
But this study only use the data from batik texture alone, do not use the color data of
batik.
In another research using the Wavelet Transform as a method for extracting
features batik and Neural Network as a method for classify batik feature said that the
highest accuracy of 100% for the testing data is the same as the training data and
78.26% is achieved for testing data is the air-different from the training data. Both
accuracy obtained on the value of learning rate 0.8, using the momentum of 0.9, the
number of hidden layer nodes composition [40 10 1] on the 5th level decomposition.


These results are explained by Bernardine Arisandi, Nanik Suciati, and Arya Wijaya
Yudhi in a paper entitled Pengenalan Motif Batik dengan Rotated Wavelet Filter
dan Neural Network [6]. However, as in previous papers, the research is done by
taking shape features of batik, does not include other features such as color features.

I. WORK SCHEDULE

No. Detail Month
A. PREPARATION I II III IV V VI
1 Establishment of Work Plan
3 Data Gathering
4 Tools Preparation
B. IMPLEMENTATION
1. Design Flowchart
2. Design Flow of Program
3. Design Flow of Database
4. Design Interface (GUI)
5. Making Application
6. Testing Application
7. Bug and Error Correction
C. MAKING REPORT
1. Data Analysis
2. Writing Draft Report
3. Revision

J. COST ESTIMATE
Cost of Tools
No Tools Name Use for Cost
1 Smartphone Android Device used for research Rp. 4.000.000
2 Camera Device used for capturing ba-
tik image for dataset
Rp. 4.000.000
3 Paper 1 rim Print paper Rp. 60.000
4 Transportation fee Transportation for searching
data batik
Rp 100.000



K. REFERENCES
[1] Barakbah, A., Kiyoki, Y.: 3D-Color Vector Quantization for Image Retrieval
Systems
[2] http://en.wikipedia.org/wiki/Image_moment
[3] Barakbah, A., Kiyoki, Y.: Image Search System with Automatic Weighting
Mechanism for Selecting Features
[4] Pratama, Alvian A., Suciati, N., Purwitasari, D.: Implementasi Fuzzy C  Means
untuk Pengelompokan Citra Batik Berdasarkan Motif dengan Fitur Tekstur
[5] Pratikaningtyas, D., Santoso, I., Ajulian Z, A.: Klasifikasi Motif Batik
Menggunakan Metode Transformasi Paket Wavelet
[6] Arisandi B., Suciati N., Wijaya A.: Pengenalan Motif Batik dengan Rotated
Wavelet Filter dan Neural Network.