ibul 2004
Feal Part
1 magi nary
Part
i t ude
x wavelet
lution is 1
'esolution
groups of
registered
different
iss filter,
discrete
e used to
. We find
resolution
e spatial
ave been
spectral
original
g (d), (e)
e original
is similar
, but (f) is
ransform,
wavelet
ve can put
n method
€ quality
, standard
mensions.
grayscale
easure the
ds shows
finition of
ibundance
el value in
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
the image. The statistics is shown in tab 1.
Table 1 shows that average value of fig 8(d) is closest to fig 7(b),
fig 8(e) and fig 8(f) are very close to fig 7(b), but fig 8(c) is
greatly different from fig 7(b). This demonstrates that the
spectral characteristic of fig 8 (d) is closest to fig 7(b), the
spectral characteristics of fig 8 (e) and fig 8 (f) are similar with
fig 7(b). There is evident spectral distortion in fig 8(c). The
conclusion accords with our observation. The statistics of
standard difference and entropy show that spatial resolution of
all the fused images have been improved, fig 8(c) is the clearest,
fig 8 (f) is next best. Average grads and fractal dimensions of fig
8 (f) both are highest, since the details have been enhanced in
fusion process, roads, bridges, airports, rivers and other objects
are distinguished more easily.
Table 2 shows the result similar with table 1. Standard
difference, entropy, average grads and fractal dimensions of fig
10 (f) all are highest, which shows obviously the details are
enhanced. But the average value of fig 10 (f) is higher than other
fused images except fig 10 (e), which demonstrates some
distortions exist.
In a word, though there is slight spectral distortion in the fused
image based on complex wavelet transform, it's spatial
resolution and details texture have been enhanced remarkably.
This demonstrates that the fusion method based on complex
wavelet transform is better than the fusion method based on
discrete wavelet transform and discrete wavelet packet
transform.
6. CONCLUSIONS
In this paper, first we introduce a dual-tree complex wavelet
transform with approximate shift invariance, good directional
selectivity, PR, limited redundancy and efficient computation.
Then we carry out image fusion using CWT instead of classical
DWT, design a image fusion approach based on CWT.
Experiment results show that the fusion method based on CWT
is better than the fusion method based on DWT and DWPT.
REFERENCES
A. Jalobeanu , L. Blanc-Féraud , J. Zerubial 12000, Satellite
image deconvolution using complex wavelet packets Thème
3— Interaction données,
connaissances Projet Ariana Rapport de recherche n? 3955
—116 pages
homme-machine,images,
Javier Portilla and Eero P. Simoncelli, 1999, Texture modeling
and synthesis using joint statistics of complex wavelet coeffici-
ents, Proceedings of the IEEE Workshop on Statistical and
Computational Theories of Vision, Fort Collins, CO,
http://www.cis.ohio-state.edu/ szhu/SCTV99.html
JIANG Han-ping, WANG Jian-min, 2000, Application of
Complex-valued Wavelet Transform to Image Matching,
Journal of Jiang-xi Institute of Education (Natural Science),
Vol.21 No.6, pp.29-31
Julian Magarey and Nick Kingsbury! 1998, Motion estimation
using a complex valued wavelet transform |/EEE Trans. on
Signal Processing, special issue on wavelets and filter banks,
vol 46, no 4, pp 1069-84.
N. Kingsbury, 1998a, The dual-tree complex wavelet transform:
A new technique for shift-invariance and directional filters, in
DSP Workshop
Nick Kingsbury © 1998b, The dual tree complex wavelet
transform: A new efficient tool for image restoration and
enhancement( Proc. European Signal Processing Conference,
EUSIPCO 98, Rhodes, pp 319-322.
Peter de Rivaz and Nick Kingsbury 12001, Bayesian image dec-
onvolution and denoising using complex wavelets! Proc. IEEE
Conf. on Image Processing, Greece, Oct 8-10, paper 2639.
Serkan Hatipoglu, Sanjit K. Mitra, and Nick Kingsbury, 1999,
Texture classification using dual-tree complex wavelet
transform, /EE. Image Processing and its Applications,
Conference Publication No.465
S G Mallat, 1989, A theory for multiresolution signal
decomposition: The wavelet representation, /EEE Trans. PAMI,
11(7), pp674-693
Average Value Standard Difference Entropy Average Grads Fractal Dimensions
SPOT panchromatic image 130.327910 40.038097 6.987931 12.564874 2.952943
Landsat7 TM multi-spectral image 128.069414 23.931238 11.223637 4.046452 2.907271
direct power average 123.700075 21.678290 10.764350 5.870725 2.943095
high pass filter 123.699798 21.678528 10.764387 5.870922 2.943089
IHS transform 126.661916 43.560041 12.659949 13.679566 2.947926
discrete wavelet transform 127.556114 26.931699 11.492410 10.465421 2.956894
discrete wavelet packet transform 136.432717 24.632584 11.243407 8.922076 2.962865
complex wavelet transform 136.289744 28.501428 11.467674 12.921713 2.986830
Table 1. Statistics of quality indexes for evaluating the images in fig. 7 and fig. 8
533