TELKOMNIKA ISSN: 1693-6930
Fusion of Infrared and Visible Images Based on NSCT Tingman Zhang 617
No. 1
2 3
4 5
6 Original Image
Low-Frequency NSCT Coefficient
Low-Frequency Coefficient PC Diagram
Low-Frequency Coefficient GWPC
Diagram
Figure 4. Image phase congruency diagram of different structure features
3.2. High-Frequency Coefficient Fusion Based on Gauss Weight SML
The high-frequency coefficients decomposed by NSCT represent the detail information of infrared and visible images. The fusion method to choose the maximum absolute value
doesn’t take the impact of the neighborhood pixels into consideration and as an operation method based on points, this method is easy to be affected by noises. Sum Modified Laplacian
SML better reflects the edge features of different image structures and it can represent the image sharpness and definition. This paper fuses high-frequency coefficients via Gauss weight
SML, which is defined as follows:
2 ,
,
, ,
Q P
l k l k
a P b
Q
SML i j
ML i
a j b
12
, ,
, ,
, ,
,
, 2
, ,
, 2
, ,
,
l k l k
l k l k
l k l k
l k
ML i j
H i j
H i
r j H
i r j
H i j
H i j
r H
i j r
13 Here,
, l k
H is the 1st-order k-direction high-frequency sub-band coefficient after NSCT
decomposition and r is the radius of the neighborhood window. Formula 12 doesn’t take the influence of the pixel position into account. Literature [13] proposes a normalized SML NSML,
introduces position information into ML through weighting matrix and emphasizes the contribution of the pixels at different distances to the central pixel. This literature uses the
following weight formula:
2 1 2 Q 1
2 1 2
1
1 ˆ
, ˆ ,
P P
Q P
Q a
P b
Q
a b W
W a b ω
14
2 1 2 Q 1
ˆ
T P
W X Y
15
TELKOMNIKA Vol. 14, No. 2
618 Here,
[1, 2, ,
X P
,
dimensional diagram of its we weight calculation method is n
central pixel is emphasized a weight. The sum of the norma
is as follows:
2 1 2
1
,
P Q
a b e
ω
Here, σ
is the varianc
a Weig
Gauss weight SML is d
,
, ,
Q P
l k a
P b Q
WSML i j
a b ω
Here, weight function high-frequency coefficients of
and the fused coefficient and w
, ,
l k A
l k B
F WSML
A WSM
H W
H W
, ,
, l k
A A
WSML l k
l k A
B
WSML W
WSML WSML
,
4. Experiment Result and An 4.1. Objective Evaluation Crit