IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, VOL. 10, NO. 3, MAY 2013
I. I NTRODUCTION
Manuscript received February 21, 2012; revised June 17, 2012; accepted
July 15, 2012. Date of publication September 6, 2012; date of current version November 24, 2012. This work was supported by the Korea Aerospace
Research Institute.
J. Choi is with the School of Civil Engineering, Chungbuk National University, Cheongju 361-763, Korea (e-mail: jaewanchoi@chungbuk.ac.kr).
J. Yeom, A. Chang, and Y. Kim are with the Department of Civil and
Environmental Engineering, Seoul National University, Seoul 151-742, Korea.
Y. Byun is with the Satellite Information Research Center, Korea Aerospace
Research Institute, Daejeon 305806, Korea.
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/LGRS.2012.2210857
(1)
491
(2)
(3)
(4)
Fig. 1. High-frequency information of hybrid pansharpening. (a) Panchromatic image, (b) primary high-frequency information of the blue band, and
(c) secondary high-frequency information of the blue band.
492
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, VOL. 10, NO. 3, MAY 2013
TABLE I
DATA S PECIFICATIONS
MSln
(I l )
,
,
n = max
(I l ) MSln
N
if
n=1
1
N entropy
entropy(I l )
N
l
MSln
(I )
, if
MSln
n=1
thr
1
N entropy
MSln
entropy(I l )
> thr
(6)
(7)
493
TABLE II
C OMPARATIVE Q UICK B IRD /KOMPSAT-2/G EOEYE -1 PANSHARPENING R ESULT
Fig. 3. Detail of 300 300 of QuickBird. (a) Multispectral image and the
(b) GIHS, (c) GSA, (d) GSA-CA, (e) AIHS, and (f) hybrid results.
Fig. 4. Detail of 300 300 of KOMPSAT-2. (a) Multispectral image and the
(b) GIHS, (c) GSA, (d) GSA-CA, (e) AIHS, and (f) hybrid results.
That means that D and Qavg are more efficient indexes than
the ERGAS/RASE to estimate spectral quality. In the case
of the KOMPSAT-2 data set, the AIHS result had the best
spectral quality index value among the algorithms. In addition,
AIHS methods had the least AG value, and the fused image
by the AIHS of KOMPSAT-2 did not virtually include spatial
information of panchromatic imagery such as that in Fig. 4(e).
Therefore, our hybrid pansharpening result has better spatial
494
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, VOL. 10, NO. 3, MAY 2013
IV. C ONCLUSION
In this letter, a hybrid pansharpening algorithm for improving
the spatial quality has been developed. By injecting highfrequency information of two types and postprocessing the
fusion parameter, we have empirically optimized the spectral
and spatial qualities of pansharpened images. From the experiments, the images obtained with our hybrid pansharpening had
better spatial sharpness and clarity than with the GIHS and
AIHS methods while retaining spectral characteristics closer
to the original multispectral image than those of the GS-based
methods. Through evaluation indexes and visual comparison,
the capability of our algorithm has been confirmed.
R EFERENCES
Fig. 5. Detail of 300 300 of Geoeye-1. (a) Multispectral image and the
(b) GIHS, (c) GSA, (d) GSA-CA, (e) AIHS, and (f) hybrid results.
sharpness than the other methods while overcoming the tradeoff problem between the spectral preservation and the spatial
sharpness of CS-based algorithms. As shown in Figs. 35, our
hybrid result has better spatial sharpness than the other methods
while maintaining color information visually similar to the
original multispectral image [see Figs. 3(f) and 4(f)]. In the case
of other fusion results, some color distortion occurred, such as
that in Figs. 3(b), 4(b), and 5(b). In addition, some blurring
occurred in some of the features, such as building edges and
moving cars [see Figs. 3(c)(e), 4(c)(e), and 5(c)(e)]. These
results show that hybrid pansharpening can improve the spatial
quality dramatically while preserving the spectral information
of the original multispectral image.