104 CHINESE OPTICS LETTERS / Vol. 6, No. 2 / February 10, 2008
A novel image fusion method using WBCT and PCA
Qiguang Miao (
¢¢¢
ééé
222
)
1,2
and Baoshu Wang (
äää
)
1
1
School of Computer Science, Xidian University, Xi’an 710071
2
Information & Communication College, Guilin University of Electronic Technology, Guilin 541004
Received May 29, 2007
A novel image fusion algorithm based on wavelet-based contourlet transform (WBCT) and principal com-
ponent analysis (PCA) is proposed. The PCA method is adopted for the low-frequency components. Using
the proposed algorithm to choose the greater of the active measures, the region consistency test is per-
formed for the high-frequency components. Experiments show that the proposed method works better in
preserving the edge and texture information than wavelet transform method and Laplacian pyramid (LP)
method do in image fusion. Four indicators for the fusion image are given to compare the proposed method
with other methods.
OCIS codes: 100.7410, 100.0100, 100.2000.
Image fusion can process the images obtained from
different sensors by a specific algorithm so that the
resultant image is more reliable, clearer and more
intelligible
[1,2]
. In recent years, multi-scale and multi-
resolution fusion methods are widely used in image fu-
sion field and many research achievements are gotten, for
example, wavelet transform method, Laplacian pyramid
(LP) method
[3]
.
In capturing the geometry of image edges, however,
there are limitations of the commonly used separable
extensions of wavelet transforms. Two-dimensional (2D)
wavelets are good at isolating the discontinuities at edge
points, but do not “see” the smoothness along the con-
tours. Therefore, several transforms have been proposed
for image signals, which have incorporated directionality
and multiresolution and hence, could capture edges in
natural images more efficiently. Steerable pyramid
[4]
,
curvelet
[5]
and contourlets
[6]
are some popular examples.
The contourlet transform is one of the new geometrical
image transforms, which can efficiently represent images
containing contours and textures
[6,7]
. Unfortunately,
the contourlet transform is a redundancy transform be-
cause LP is employed in the first stage. Eslami and
Radha proposed a new non-redundant image transform,
the wavelet-based contourlet transform (WBCT)
[8,9]
.
WBCT is more appropriate for the analysis of the signals
which have line or hyper-plane singularity than wavelet,
and it has better approximation precision and sparsity
description. When introducing WBCT to image fusion,
we can take the features of original images well and pro-
vide more information for fusion. The fused image could
represent almost the same detail as the original image
because WBCT represents edges better than wavelets.
Principal component analysis (PCA)
[10]
is a common tool
for image enhancement and data compression. In this
paper, we propose a new image fusion method combing
WBCT and PCA.
The block structure for WBCT filter bank is shown
in Fig. 1 and an example of its frequency partition is
shown in Fig. 2. In WBCT, the wavelet transform (WT)
captures the point discontinuities, and a direction filter
band (DFB) links point discontinuities into linear struc-
tures. WBCT involves basis functions that are oriented
at 2
N
(N = 1, 2, 3, · · · ) directions with flexible aspect
ratios. From Fig. 2 we can see that the frequency is
divided, where the number of directions is increased
with frequency. With such richness in the choice of ba-
sis functions, WBCT can represent any one-dimensional
(1D) smooth edges with nearly optimal efficiency.
Figure 3 is the decomposition structure of WBCT with
3-scale wavelet decomposition. The direction numbers of
DFB are 8, 4, 4. The decomposed image of Barbara is
shown in Fig. 4.
The fusion framework using WBCT is shown in Fig. 5.
First, the source images are decomposed using WBCT.
Suppose that the original images are A and B, the
Fig. 1. Framework of WBCT.
Fig. 2. Frequency partition.
1671-7694/2008/020104-04
c
2008 Chinese Optics Letters