Ϯϵ
ƚŚ
Ed/KE> Z/K ^/E KE&ZE
;EZ^ ϮϬϭϮͿ
Ɖƌŝů ϭϬ Ǧ ϭϮ ϮϬϭϮ &ĂĐƵůƚLJ ŽĨ ŶŐŝŶĞĞƌŝŶŐĂŝƌŽ hŶŝǀĞƌƐŝƚLJ ŐLJƉƚ
C14. Estimation of the Optimal Set of Parameters for PAN-Sharpening of Satellite
Images Based on the Non-Sub-sampled Contourlet Transform
Mohamed R. Metwalli
1
, Ayman H. Nasr
1
, Osama S. Farag Allah
2
, S. El-Rabaie
2
, and Fathi E. Abd El-Samie
2
1
Data Reception, Analysis and Receiving Station Affairs Division, National Authority for Remote Sensing and
Space Sciences, 23 Joseph Broz Tito st., El-Nozha El-Gedida, Cairo, Egypt
2
Faculty of Electronic Engineering, Menoufia University, Menouf, 32952, Egypt
ABSTRACT
Recent studies show that hybrid PAN-sharpening methods using the Non-Sub-sampled Contourlet Transform
(NSCT) and classical PAN-sharpening methods like the Intensity, Hue and Saturation (IHS), Principal Component
Analysis (PCA), and Adaptive Principal Component Analysis (APCA), reduce the spectral distortion in the PAN-
sharpened images. The NSCT is a shift-invariant multi-resolution decomposition. It is based on Non-Sub-sampled
Pyramid (NSP) decomposition and Non-Sub-sampled Directional Filter Banks (NSDFB). We compare the
performance of the APCA-NSCT using different NSP filters, NSDFB filters, number of decomposition levels, and
number of orientations in each level on Spot4 data with spatial resolution ratio 1/2, and QuickBird data with
spatial resolution ratio 1/4. Experimental results show that the quality of PAN-sharpening of remote sensing
images of different spatial resolution ratios using the APCA-NSCT method is affected by NSCT parameters. For
the NSP, the ‘maxflat’ filters have the best quality. For NSDFB the ‘sk’ filters have the best quality. Changing the
number of orientations in the same level of decomposition in the NSCT has a small effect on both the spectral and
spatial quality. The spectral and spatial quality of PAN-sharpened images mainly depends on the number of
decomposition levels. Too few decomposition levels result in poor spatial quality, while excessive levels of
decomposition result in poor spectral quality.
Keywords: NSCT, PCA, APCA, PAN-sharpening.
I. INTRODUCTION
Currently, several remote sensing image fusion methods have been developed, including the frequently
used Brovey method, IHS method, PCA method, and the recently developed Multi-Resolution Analysis (MRA)
techniques. The MRA techniques include the Wavelet Transform (WT), filter banks, Laplacian pyramids, and
redundant multi-resolution structures, such as the Undecimated Discrete Wavelet Transform (UDWT) and the “A
Trous” Wavelet Transform (ATWT) [1]. Multi-resolution algorithms have the ability to apply hierarchical
decomposition of an input image into successive coarser approximations. Such decomposition separates low-
frequency components from high-frequency components in both the PAN and Multi-Spectral (MS) images in a
scale-by-scale manner. Thus, the low-frequency components of MS images can be used without any modifications
during the fusion process, and consequently, spectral distortion is limited. The missing spatial information in the
MS image can be inferred from the high-frequency components of the PAN image [2]. The redundant multi-
resolution decompositions do not require sharp digital filters and are not critically sub-sampled. The typical
injection artifacts appearing in the images fused by means of conventional wavelet analysis, like ringing effects
and canvas-like patterns disappear, when the images are fused using the redundant MRA. Therefore, the
redundant MRA methods are particularly suitable for PAN-sharpening [1].
The Contourlet Transform (CT) provides a new representation system for image analysis. The CT is so
called because of its ability to capture and link the discontinuity points into linear structures (contours). The two
stages used to derive the CT coefficients involve a multi-scale transform and a local directional transform. The CT
provides 2
l
directions at each scale, where l is the number of required orientations. The CT lacks the shift-
invariance and causes pseudo-Gibbs phenomena around singularities. The NSCT is a shift-invariant version of the
CT, and it is based on the NSP and the NSDFB. Fusion based on the NSCT can use several methods for the
injection of the image details. A number of hybrid methods have been developed to combine the best aspects of
classical methods and multi-resolution transforms. Shah et al. [3] presented PCA-CT PAN-sharpening of high
resolution (Ikonos and QuickBird) and medium resolution (Landsat7 ETM+) datasets. The NSCT provided better
results than the sub-sampled CT.
Xiaohui et al. [4] proposed an intensity component addition technique based on IHS transform and
NSCT to preserve spatial resolution and color content. The IHS transform is applied on the MS image first, then
the directional sub-band decomposition of the NSCT of the PAN image is added to the directional sub-band
decomposition of the NSCT of the intensity component (I). Experiments showed that this method can improve
spatial resolution and keep spectral information, simultaneously. Shah et al. [5] presented a combined adaptive
PCA–NSCT approach for PAN-sharpening, which uses NSCT for the spatial transformation to capture intrinsic
271 978-1-4673-1887-7/12/$31.00 ©2012 IEEE