The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
(g) normal wavelet transform fused image (h) the novel method fused image
Figure. 2 SAR image, multispectral image, and the result images of different fusion methods
4.1 Subjective Qualitative Analysis and Comparison
As for visual effect, the normal wavelet transform and the novel
method have the advantages of preserving spectral information
over the other methods. Brovey is the simplest, but the fusion
effect is the worst. Triangle IHS, PCA and Gram-Schmidt keep
too much information of SAR image, but lose most of spectral
information. Besides, the spectral information distributes
unevenly, resulting the distortion of the image hue. In so far as
the normal wavelet transform and the novel method, the former
is not as good as the latter in preserving information of the SAR
image, and also with hue distortion from the original
multispectral image; while the novel method not only preserves
spectral information better, but also auto-adaptively keeps the
texture feature of the SAR image, so that the fused image is
smooth and natural in visual. Furthermore, the novel method
has the advantage of preserving information of SAR image. For
instance, the representative object circled by the red rectangle
in
In statistics, mean ju and standard deviation o are
defined:
¿=—( 6 )
n m
~ 2
(J
(7)
Figure 1(a) is the special information in the SAR image, which
preserved fully in the novel method, while relatively poor for
PCA or normal wavelet transform method. Comprehensively,
concerning visual effect, the novel method is relatively the best
fusion method.
4.2 Objective Quantitative Analysis and Comparison
Certain objective quantitative analyzing indexes are utilized to
compare the fusion effects and information preservation
abilities of different fusion methods adopted in this paper. We
employed mean, standard deviation, entropy, correlation
coefficient, average gradient and spectral distortion to reflect
image luminance, spatial detail information preservation, and
spectral information preservation (Luclen, 1997).
1. Mean and Standard Deviation (S. D.):
where, n is the sample size, x. is the value of the z'th
sample.
Here, mean is the average gray value of all the pixels in
the image, and for the human visual system, it is the
average luminance. So if the mean is moderate, the visual
effect is fine. Standard deviation represents the discrete
degree of the pixel value relative to the mean. The larger
the deviation is, the more dispersedly the gray grades
distribute, and if the probability of each gray grade equals,
the information capacity is the maximum.
1132