Adaptive remote sensing image fusion technique Essay

REMOTE SENSING IMAGE FUSION
The main challenge was obtaining suitable weights by the
adaptive model. This paper proposed an adaptive remote sensing image fusion technique based on the DCT with adaptive
weight of the fusion rule based on the PSO. It is illustrated as
on Fig.1
Fig. 1. The scheme of the adaptive remote sensing image fusion technique
based on the DCT and PSO .
A. The adaptive weight fusion rule based on the PSO
In this section, the adaptive weighted of fusion rule based
on the PSO is described. The fusion rule based on the
weighted averaging is a simple image fusion technique. The
average of the source images is taken pixel by pixel as the
following equation
F(x,y) = w A(x,y)+(1 w) B(x,y) (3)
Where F(x,y) is the fused image, A(x,y) and B(x,y) are coefficients of the DCT of the source images. The W and (1W)
are scalar weights in the range [0,1]. The weighted averaging
technique is easily implemented and fast executed. But the
weighted averaging approach causes noises with the existing
in the source images, it causes suppresses salient features that
should be preserved for the fused image, producing a low
contrast result. To overcome these disadvantages and get better
results the PSO was used to optimize the wighted W in eq.3.
Algorithm 1 The adaptive the weighted of fusion rule based
on PSO
1 Input the paramters of the PSO
Population size =50
No. of iterations =100
Velocity limitation(Vmax) =17.
C1=0.1.3 and c2=0.1.
2 Initialized particles position and velocity vector random.
3 Compute fitness function for each particles (i)
4 Get the particle best (pbest)
5 Obtain the Global best (gbest)
6 Calculate particle velocity eq. 1
7 do untill No. of iteration =100
8 Obtain the optimal solution for w in eq. 3
B. The adaptive remote sensing image fusion based on the
DCT technique.
This paper proposed an adaptive remote sensing image
fusion technique based on the DCT and PSO. The proposed
technique combines the PCA transform and the DCT to
achieve the fusion process and the PSO is used to optimize the
weight of the fusion rule. The two source images MS and Pan
image are used as input images. At the first the two images
were registered perfectly and the MS image was resampled
then the PCA was used. The PCA transform reduced the data
and extracted the features. It converted a set of multi-spectral
bands three or more (MS) image to principal components. In
this case, there were three PCs. The PC1 has the most spatial
information of the MS image. Then the DCT transform used
the PC1 and Pan image as the inputs. The DCT coefficients
are computed for each image. The DCT coefficients focused
in the low-frequency region. Then the fusion rule after the
wighted optimized was applied to get the fused coefficients.
In this proposed technique, the fused image is obtained by
applying the inverse of the DCT transform (IDCT) and the
inverse of the PCA transform (IPCA)