Fusion of multispectral and panchromatic images based on support value
transform and adaptive principal component analysis
Shuyuan Yang
⇑
, Min Wang, Licheng Jiao
Key Laboratory of Intelligent Perception and Image Understanding of Ministry of Education of China, National Key Lab of Radar Signal Processing,
Department of Electrical Engineering, Xidian University, Xi’an 710071, China
article info
Article history:
Received 24 January 2010
Received in revised form 11 September 2010
Accepted 21 September 2010
Available online 29 September 2010
Keywords:
Fusion
Support value transform
Adaptive PCA
ARSIS
Texture extraction
abstract
In this paper we combined the projection–substitution with ARSIS (French acronym for ‘‘Amélioration de
la Résolution Spatiale par Injection de Structures”, i.e., Improving Spatial Resolution by Structure Injec-
tion) concept assumption for fusion of panchromatic (PAN) and multispectral (MS) images. Firstly sup-
port value filter (SVF) is used to establish a new multiscale model (MSM), support vector transform
(SVT), and adaptive principal component analysis (APCA) is then employed to select the principal com-
ponents of MS images by means of a statistical measure of the correlation between MS and PAN images;
secondly, a local approach is used to check whether a structure should appear in the new principal com-
ponent and PAN high frequency structures are transformed by high resolution interband structure model
(HRIBSM) before inserting in the MS modalities. Because SVT is an undecimated, dyadic and aliasing
transform with shift-invariant property, the fused image can avoid ringing effects suffered from sam-
pling. Additionally, the ARSIS concept can make full use of the remote sensing physics to reduce the spa-
tial and spectrum distortion in the structure injection. Texture extraction is also employed to avoid the
spectral distortion caused by the mistaken injection of low-pass components into the MS images. Exper-
imental results including visual and numerical evaluation also proves the superiority of the proposed
method to its counterparts.
Ó 2010 Elsevier B.V. All rights reserved.
1. Introduction
The fusion of low resolution multispectral (MS) and high reso-
lution panchromatic (PAN) images is a useful technique for
enhancing the spatial quality of MS images [1]. A variety of im-
age-fusion techniques have been devoted to merge MS and PAN
images which exhibit complementary characteristics of spatial
and spectral resolutions. These classical methods can be classified
into three categories: statistical technique based methods, mathe-
matical technique based methods and intensity based methods.
The statistical technique based approaches convert inter-
correlated multispectral bands into a new set of uncorrelated com-
ponents by statistical tools such as principal components analysis
(PCA), and the higher resolution image is used to replace the calcu-
lated principal component [4]. The mathematical technique based
fusion methods are based on the mathematical combination of the
multispectral images and high resolution PAN image such as
Brovey method [2]. Its basic idea is that each normalized multi-
spectral image is multiplied by the PAN image to add the spatial
information to the fused image. The third category modify the
PAN image to look more like the intensity component, with replac-
ing the high frequency part of the intensity component with that
from the PAN image [3,4]. Some variation of these classical fusion
methods are also discussed, for example, Tu et al. [5] presented a
fast IHS fusion approach with spectral adjustment for IKONOS data,
later Choi [6] introduced a tradeoff parameter in fast HIS, and
Alparone [7] proposed a generalized intensity modulation way by
setting a threshold to modulate the intensity component. In recent
years, there is an increasing of interests in this field [8–13], for
example, Wang et al. [8] proposed a comprehensive framework
for general fusion of MS and PAN images; Fasbender [9] suggested
an approach within a Bayesian framework; Otazu et al. [10] consid-
ered the physical electromagnetic spectrum responses of sensors
during the fusion process for better fusion result; Garzelli et al.
[11] proposed an optimum algorithm in the minimum mean-
square-error (MMSE) sense, whose solution can minimize the
squared error between the MS images and the fusion result ob-
tained by spatially enhancing a degraded version of the MS images.
For the fusion of PAN and MS images, a general framework is to
sharpen low resolution MS image by injecting high-pass details ta-
ken from the higher resolution PAN image [14,15], namely, projec-
tion–substitution approach. Some multiresolution analysis tools
like wavelets [16–18] and pyramids [19] have been used to reduce
1566-2535/$ - see front matter Ó 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.inffus.2010.09.003
⇑
Corresponding author. Tel.: +86 029 88204298; fax: +86 029 88201023.
E-mail address: syyang@xidian.edu.cn (S. Yang).
Information Fusion 13 (2012) 177–184
Contents lists available at ScienceDirect
Information Fusion
journal homepage: www.elsevier.com/locate/inffus