Image Recognition and
Processing Using Artificial
Neural Network
Md.
Iqbal
Quraishi
, J Pal
Choudhury
and
Mallika
De, IEEE
Outline
Introduction
Related work
Method
Experiments Result and Analysis
Conclusion
2
Introduction
Artificial Neural Networks may be
considered as much more
powerful
because it can solve problems where how
to solve have been not known exactly.
Uses of artificial neural network have
been spread to a wide range of domain
like
image recognition
,
fingerprint
recognition
and so on.
3
Related work(1/4)
The appearance of digital computers and
the development of
modern theories of
learning
and
neural processing
both
occurred at about the same time, during
the
late 1940s
.
To model individual neurons as well as
clusters of neurons, which are called
neural networks.
4
Related work(2/4)
A new approach for
feature extraction
based on the calculation of
eigen
values
from a contour was proposed and found
that using
feed forward neural network
satisfactory results were obtained.
5
Related work(3/4)
6
Feed Forward Neural Network
Related work(4/4)
7
Method
Processing of Original Image
◦
The initial optimal image has been taken as
furnished in Fig

2 which has been considered
as original image.
8
Fig

2
Table

1
Input Data Matrix
Method
The average error after insertion of
salt
and pepper noise
has been calculated
which is 25.67%.
9
Table

2
Input Data Matrix with Noise
Fig

3
Method
Processing of Noisy Image
◦
Adaptive median Filter
has been applied on
noisy image such that the noise can be
removed and the output image would be
considered as filtered Image.
◦
The
estimated Error
and
average error
of the
values stored in filtered image matrix have
been calculated with reference to the values
stored in original data matrix. The a
verage
error
has been found as
5.397%
.
10
Method
The original image after removal of noise
has been transformed into data matrix
containing pixel values which have been
furnished in Table

3.
11
Table

3
Input Data Matrix after Noise
Removal
Fig

4
Method
For easier calculation
four pixels
have
been taken together.
The binary values of four pixels together
side by side have been combined and
formed as
32 bit binary number
.
Now the 32 bit binary number has been
converted into a decimal number.
12
Method
The decimal number as generated in
page
11
has been placed in original data matrix
termed as
ORMAT[][]
.
13
Table

4
Original Data Matrix ORMAT[][]
Method
The instructions furnished in
page 12
to
page 13
have been repeated for the total
pixel value of the original image after
noise removal as stored in Table

3.
Therefore a matrix has been produced
which has been stored in data matrix
termed as
ORMAT[][]
as furnished in
Table

4.
14
Method
Processing of second Image(Test Image)
◦
A new image has been taken which is
considered as a test image.
◦
Now it is
necessary
to check whether the
said image can be recognized or not.
15
Fig

5
Table

5
Test Data Matrix
Method
Instructions as furnished in
page 9
have
been executed on test image to generate
test data matrix with noise as furnished in
Table

6.
16
Fig

6
Table

6
Test Data Matrix with Noise
Method
Instructions as furnished in
page 10
have
been executed on test image with noise
to generate test data matrix after noise
removal as furnished in Table

7.
17
Fig

7
Table

7
Test Data Matrix after Noise Removal
Method
Procedures as mentioned from
page 11
to
page 13
have been executed on test
image after noise removal to generate the
decimal number which has been placed in
test data matrix
TESTMAT[][]
.
18
Table

8
TESTMAT[][]
Method
Calculation of Average Error of test data
matrix based on original data matrix.
◦
The estimated error and average error of the
values stored in decimal matrix as furnished in
Table

9
have been calculated with reference
to the values stored in original data matrix as
stored in
Table

4
. The
average error
has been
found as
31%
.
19
Method
20
Since the average error is
less than 45%
,
necessary steps regarding the processing
of test image has been made using the
technique of artificial neural network for
the purpose of recognition.
Table

9
Estimated Error Data
Method
Processing of Image towards recognition using
Artificial Neural Network.
◦
The
feed forward back propagation neural
network
has been used on the test data
matrix of the test image for training and
testing with reference to the original data
matrix of the original image.
◦
A new data matrix named NEWMAT[][] has
been produced as a result which has been
furnished in Table

10.
21
Method
It takes considerably
less time
to
complete the training and Testing using
ANN.
22
Table

10
Data Matrix NEWMAT[][] after ANN
application
Method
Each value of the data matrix
NEWMAT[][]
has been converted
into 32
bit
binary number.
Now the 32 bit binary number has been
divided into four 8 bit binary numbers.
Each 8 bit binary value has been
converted into decimal and each of them
has been considered as pixel values for
four consecutive pixels row wise.
23
Method
The instructions furnished in
page 23
have
been repeated for the total values of the
data matrix
NEWMAT[][].
As a result a new modified data Matrix
named
MODMAT[][]
has been produced
as furnished in Table

11.
24
Method
25
Table

11
Modified Data Matrix MODMAT[][]
Fig

8
Method
26
Calculation of estimated Error and Average
Error.
◦
The estimated error and average error of the
values as stored in Table

11 with reference to
the values stored in Table

3 have been
calculated and the
average error
has been
found as
14.39%
.
Experiments Result and Analysis
27
Serial
Number
Original Image
Noisy Original
Image
Average Error
with
respect to Original Image
1
25.67%
2
26.42%
Experiments Result and Analysis
28
Serial
Number
Original Image
after
Noise Removal
Average Error
with
respect to
Original Image after Noise Removal
1
5.39%
2
2.93%
Experiments Result and Analysis
29
Serial
Number
Test Image
Noisy Test
Image
Average Error
due to Noise
with
respect to Test Image
1
25.75%
2
27.39%
Experiments Result and Analysis
30
Serial
Number
Test Image
After
Noise Removal
Average Error
with
respect to Test Image
1
5.56%
2
7.8%
Experiments Result and Analysis
31
Serial
Number
Average
Error
with
respect to
Original Image
after Noise
Removal
Test Image
after
training using
ANN
Average Error
with
respect to
Original Image
Remarks
1
31%
14.39%
Recognition
Possible
2
64%


Recognition
Not
Possible
Conclusion
If the average error is
less than 45%
,
Artificial Neural network
can be applied
for training and testing for the purpose of
recognition.
Therefore the test image is recognized
and matched successfully with original
image.
32
Conclusion
If the average error is
greater than 45%
then the image is recognized as a different
image.
It takes less time for training and testing
using ANN as number of rows of the
matrix used for training has
one fourth
number
of columns compare to the
original image.
33
Thank you for you listening
34
Comments 0
Log in to post a comment