The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
1157
registered strictly at the same scale. We fused the images with
different methods including PCA and astrous. These images were
used to be compared with the image fused by PCA+astrous.
4.2 Visual and Quantitative Evaluation
4.2.1 Visual Evaluation: First the fused images were
evaluated visually. It is clearly seen from the pictures in Figure 2
that the spatial resolutions and the spectral resolutions of the
images after the fusion are improved, in comparison to the original
images. The spectral information in the original panchromatic
image has increased, and the structural information in the original
multispectral images has also been enriched. Hence, the fused
images contain both the structural details of the higher spatial
resolution panchromatic image and the rich spectral information
from the multispectral images. In the original QuickBird images, it
is very difficult to discern some physical features like small
buildings.
But, the colors of the features in the fused images are changed.
This color distortion effect is the largest in PCA method. We find
that (c) fully conserve spatial information of high-resolution image,
but evident spectral distortion exist. Among these three methods,
astrous transformation gives the best result in terms of color
conservation. The spectral characteristics of (d) are closest to the
original multi-spectral image than other fused images, and the
spectral characteristic of (d) is similar to (b).
However, the result obtained from PCA+astrous wavelet
transform approach is superior to those two results, as the colors
of the features in original multispectral images are nearly the same
with the fusion image just based on the astrous wavlet.
The mean gradient (M.G.) reflects the contrast between the
details variation of pattern on the image and the clarity of the
image. And the correlation coefficient (C.C.) between the
original and fused image shows the similarity in small size
structures between the original and synthetic images. The
higher the value, the more similar the fused image to the
corresponding original image, which in turn indicates a good
spectral information retain in the fused results. Results in
Table shows that the astrous wavelet transform approach and
the PCA+astrous keeps over 90% of the spectral content of
all original multispectral bands, the result of the astrous is the
best and the PCA+astrous-based image fusion is only slightly
lower.
But the mean gradient of the PCA+astrous-based image
fusion method is greater than those of the astrous wavelet
based image fusion method, since the details have been
enhanced in fusion process, roads, bridges, airports, rivers
and other objects are distinguished more easily. As previously
stated, if the object of image fusion is to construct synthetic
images which are closer to the reality they represent, then the
PCA+astrous-based image fusion method meet this objective
very well. This is one of the main advantages of using the
PCA+astrous transform for image fusion.
In a word, though there is a slight spectral distortion in the
fused image based on PCA+astrous wavelet transform, its
spatial resolution and details texture have been enhanced
remarkably. This demonstrates that the fusion method based
on PCA+astrous wavelet transform is better than the fusion
method just based on asttrous wavelet transform or PCA
transform.
But the PCA+astrous wavelet gives a better spatial resolution
when compared to the results from other fusion methods. So in
this paper, compared with the fused result just from astrous wavlet
or the PCA alone, the fused result obtained from the PCA+astrous
wavelet transform has a better visual effect in Quickbird image
fusion in Figure 2.
All the above analysis suggests that the wavelet based fusion
approach provides overall the best results in the methods used
in our study. Based on the experimental results obtained from
this study, the PCA+astrous-based image fusion method is
very efficient for fusing Quickbird images. This new method
has reached an optimum fusion result.
4.2.2 Quantitative Evaluation
In general, a good fusion approach should retain the maximum
spatial and spectral information from the original images and
should not damage the internal relationship among the original
bands. So secondly we evaluate the performance of the fusion
method based on PCA+astrous wavelet transform using image
quality indexes. Based on these criteria, the indexes we selected
are average value, standard difference, entropy, average grads,
fractal dimensions and correlation coefficients. Average value can
show the distribution of the image grayscale in the rough.
Standard difference and entropy can measure the information
abundance in the image. Average grads shows exiguous contrast,
varied texture characteristic and definition of the image.
Table 1 presents a comparison of the experimental results of
image fusion using the PCA image fusion method, the astrous
wavelet-based image fusion method, and PCA+astrous method in
terms of combination entropy, the average gradient and the
correlation coefficient^.
The combination entropy (C.E.) represents the property of
combination between images. The larger the combination entropy
of an image, the richer the information contained in the image. In
Table 1, the combination entropy of the PCA+astrous-based image
fusion is greater than those of other methods. Thus, the
PCA+astrous-based image fusion method is better than the astrous
wavelet and PCA methods in terms of combination entropy.
Method
C.E
M.G
C.C
Original images
' (band 1> band2,
band3, band4)
Image fused by
4.80364
4.91436
5.01511
6.37158
7.16106
5.9237
6.2839
6.6029
7.7767
17.3162
0.62786
PCA
7.17243
17.2208
0.62657
(band K band2^
7.21094
17.6439
0.62712
band3^ band4)
7.19666
18.3797
0.6512
Image fused by
7.11928
15.8315
0.94334
astrous
7.22269
16.2062
0.94166
(band K band2^
7.25106
16.3050
0.94431
band3N band4)
7.16074
17.0117
0.89956
Image fused by
7.12787
19.8423
0.90557
PCA+astrous
7.36655
20.9268
0.90967
(band K band2>
7.36208
20.8359
0.91660
band3^ band4)
7.29436
21.4049
0.86847
Table. 1. Information statistics of Quickbird image