Download as pdf or txt
Download as pdf or txt
You are on page 1of 3

Ton Duc Thang University

Faculty of Information Technology

Numerical Analysis
Module code: 502041
Anh H. Vo

1 Introduction
This assignment will apply the basic concepts of Numerical Analysis to build a
real application for the image segmentation problem.

Figure 1: The illustration of the image segmentation by using the interpolation


polynomial function

The figure in the left hand side with the blue curve shows the polynomial func-
tion f (x) of the input image, which it is computed by the frequency of each
unique intensity, and the the red curve describes the approximation of the poly-
nomial function g(x), which it interpolate by the intensity x and the frequency of
x denoted as y. Based on g(x), we are able to find the appropriate thresholding
value θ to get the image result in the right hand side.

2 Instruction
Suppose that given arbitrary image I as a matrix which represents the intensity
at each pixel. As in Table 1. To illustrate matrix I as the polynomial function
f (x). We must compute the frequency of each unique intensity in matrix I.
In Table 2, y = n(x) which means the number of observation of x intensity in
matrix I In the next step, the values of x and y will be utilized to find the
approximation of polynomial function g(x) = a0 + a1 x + ... + an xn by using

1
Ton Duc Thang University
Faculty of Information Technology

0 1 5 3 1 10 10 7 5 7
0 1 8 3 1 2 10 7 5 10
1 1 8 3 9 2 9 7 8 6

Table 1: Example of matrix I obtains by an input image

x 0 1 2 3 4 5 6 7 8 9 10
y 2 6 2 3 0 3 1 4 3 2 4

Table 2: The values of x and y which present the intensity value and the fre-
quency of each intensity in matrix I

the interpolation methods. Besides, we also need the thresholding value θ for
segmenting regions of the input image. To do this, we must find the indexing j
that at this point the value of the second derivative of the interpolation ŷ vector
is minimized. ŷ vector is computed from g(x) based on the following equation:

ŷi = g(xi ) (1)


j = argmin{(| ŷi0 0
|) } (2)
i
j
θ= , where L means the maximum of intensity (3)
L−1
Finally, the segmentation of input image will be able to obtain by equation
below:

1 if x > θ
b(x) = (4)
0 otherwise

Remarks

0
0 2 4 6 8 10

Figure 2: The illustrate of the polynomial function f (x) of the input image

2
Ton Duc Thang University
Faculty of Information Technology

• To read an input image in MATLAB:


>> I = imread ( ’ l e n a . jpg ’ ) ;

Recall that the input image must be the grayscale image, if the input
image is the RGB image, you need to convert to the grayscale image.
>> I = r g b 2 g r a y ( I ) ;

• To compute the frequency of each intensity in MATLAB:


>> [ y , x ] = i m h i s t ( I ) ;

• The built-in functions as polyfit and polyval are possibly applied to ap-
proximate the polynomial function g(x).
• In equation (2), the abs and sort functions could use to compute the
absolute value and sort of the values.

• diff in MATLAB is utilized for computing the derivative of function

3 Requirements
Requirement 1: Write a simple program to find the approximation of the
polynomial function of the input image and then segment the image re-
gions based on the interpolation methods.
1. Find x and y variables from the input image (1 point)
2. Find g(x) function which approximates the polynomial function f (x)
in the case of using the built-in function of MATLAB programming
language (1 point)
3. Find g(x) function which approximates the polynomial function f (x)
in the case of using your functions instead of the built-in functions.
(2 point)
4. Comparison of the relative errors when the approximation of the
polynomial function g(x) with n degree is 3, 4, 5, 6, respectively. (2
point).
5. Find the index which shows the minimum value of the second deriva-
tive of ŷ. (1 point)
6. Select g(x) toward the best relative error from the previous require-
ment to show the segmentation of input image. (1 point)
Requirement 2: Write a brief report to illustrate the process following the
above requirements (2 point)

You might also like