An Image Fusion Algorithm Based on Discrete Wavelet Transform and Canny Operator - Advanced Research on Computer Education, Simulation and Modeling - page 35

ED , ED , ED are all

of the three directions and the high- frequency coefficients'

1 2 3

gained.

The energy of low-frequency coefficient's each pixel is defined as follows:

2
1

2
2

2
3

EA ( x , y ) = EA ( x , y ) + EA ( x , y ) + EA ( x , y )

(10)

The energy of high-frequency coefficient's each pixel is defined as follows:

2
1

2
2

2
3

ED ( x , y ) = ED ( x , y ) + ED ( x , y ) + ED ( x , y ) (11)

xE of the frequency domain coef-
ficients of original images and make a comparison to determine the fused low-frequency
coefficient FA and fused high-frequency coefficient FD (three directions):

E 1 x , y ) 2 ( , y )

Calculate each point's energy ,

A 1 x , y ) EA 1 x , y ) ≥ EA 2 ( x , y )

FA (12) ( x , y ) =

A 2 ( x , y )

,

EA 1 x , y ) < EA 2 ( x , y )

D 1 x , y )

,

ED 1 x , y ) ≥ ED 2 ( x , y )

FD (13) ( x , y ) =

,

D 2 ( x , y ) ED 1 x , y ) < ED 2 ( x , y )

Then, implement a consistency verification towards the frequency domain coeffi-
cient by a 3 × 3 template, which is an essential part. If the fusion coefficient of certain
region's center is selected from the transformed coefficient of the original image 1,
however, the other 5 or more than 5 fusion coefficients in this region are selected from
the transformed coefficient of the original image 2, the fusion coefficient of the re-
gion's center should be changed to the transformed coefficient of the original image 2.
As for the boundary coefficient, the consistency verification rule is altered to first make
up zeros and then compare them. Finally, the fused image can be generated adopting
inverse discrete wavelet transform(IDWT).

4 Experiment Result and Data Analysis

The image fusion algorithm has been achieved using MATLAB 7. 0. 1 platform. The
original image 1 used in the experiment is that the center is fuzzy and the vicinity is
clear, and the original image 2 is just opposite. Show as the Figure 2 below.
Make a comparison of the following methods: a traditional rule that averages low
frequency coefficients and selects absolute maximum of high frequency coefficients;
this article's new fusion method is separately utilized at the low-frequency coefficient,
the high-frequency coefficient and the low- high-frequency. The fusion results are
shown in the Figure 3. Fig.3 (a) is the result of the traditional method of average. Fig.3
(b) is the method of this article acting on the high-frequency coefficient (method of this
article 1). Fig.3 (c) is the method of this article acting on the low-frequency coefficient
(method of this article 2). Fig.3 (d) is the method of this article acting on the
low-high-frequency coefficient (method of this article 3).