Transcript Handwritten Thai Character Recognition Using Fourier

Handwritten Thai Character
Recognition Using Fourier
Descriptors and Robust CPrototype
Olarik Surinta
Supot Nitsuwat
INTRODUCTION



This research proposes the method for Thai
handwritten character recognition.
The processing is based on Thai characters on
which preprocessing have been conducted.
There are 44 Thai characters:
กขฃคฅฆง จฉชซฌญ
ฎฏฐฑฒณ
ดตถทธน
บปผฝพฟ
ภมยรลว
ศษสหฬอฮ
NCCIT05
2
INTRODUCTION
Training Scheme
Image
Pre
Processing
Feature Extraction
(FD)
C, c
Recognition Scheme
Unknown
Image
Pre
Processing
RCP Training C, c Database
Scheme
Feature Extraction
(FD)
RCP Recognition
Scheme
Output
Figure 1 Thai handwritten recognition scheme flow diagrams.
NCCIT05
3
DATA PREPROCESSING
Character-images are images of Thai
hand-written characters.
 The output will be stored in the term of
digital data by scanning. One bitmap file
with gray scale pattern and 256 levels
specifics one character.

Figure 2 A prototype character-image.
NCCIT05
4
IMAGE PROCESSING

Binarization
 Binarization
converts gray-level image to
black-white image, and to extracting the
object component from background, this
scheme will check on every point of pixel.
Figure 3 The example
of binarization scheme.
NCCIT05
5
Binarization

The individual bit bares 2 possible values:
1 refers to background and
 0 refers to object

(B)
(A)
Figure 4 The diagram of extracting
the object from the background
component in the image.
NCCIT05
(C)
6
Edge Detection
Edge detection is one of an important
image processing phases.
 This paper uses chain code technique to
detect the image’s edge. The direction has
been classified by 8 categories:

Figure 5 Chain code with 8 directions.
NCCIT05
7
Edge Detection


Once the edge of image has discovered, shown in
figure 4, the process needs to find the character
line.
The coordinate xk , yk is represented by complex
number as the formula:


uk  xk  iyk


Figure 6 coordinate xk , yk
represented in character image.
NCCIT05
8
FOURIER DESCRIPTORS



Fourier Features used to describe the edge of
the object works by identify coordinate xk , yk  ;
K = 0, 1, …, N-1 where N is any other area in the
image.
All point xk , yk  will be represents as complex
number.
Therefore, the DFT can be derived as below:
N 1
f l   uk
k 0
2 

exp  j
lk 
N 

NCCIT05
l  0,1,...,N  1
9
FOURIER DESCRIPTORS


From the above formula, coefficient vector will be
automatically calculated.
This vector fits as 1 dimension with the size of
1x10 or 1xn
Figure 7 Fourier Descriptors of
Image.
NCCIT05
10
ROBUST C-PROTOTYPES (RCP)

RCP can be determined in grouping phase
in order to estimate C-Prototypes
spontaneously, utilizing loss function and
square distance to reduce some noise.

The diagram of solving the problem by
RCP is shown in figure 8
NCCIT05
11
ROBUST C-PROTOTYPES (RCP)
Figure 8 RCP algorithm.
NCCIT05
12
EXPERIMENTAL RESULT



This research paper proposes the method for
Thai Handwritten Character Recognition using
Fourier Descriptors and Robust C-Prototype
clustering.
Recognition scheme is based on features
extracted from Fourier transform of the edge of
character-image.
the character-image is described by a group of
descriptors.
NCCIT05
13
EXPERIMENTAL RESULT

We train the system using the RCP training
scheme to find the centroid of the prototype
(44 Prototypes) and membership function.

Finally, the FD of unknown character-image is
used to perform recognition step.
In this way the experimental results of
recognition, RCP can perform with accuracy up
to 91.5%.

NCCIT05
14
Figure 9 The character images the adjustment scheme.
NCCIT05
15