Spatial enhancement of multispectral images on urban areas - Urban Remote Sensing: Monitoring, Synthesis and Modeling in the Urban Environment - page 144

TABLE 10.3 Main characteristics of GS, GSA-CA, GLP-CA and GMMSE algorithms. β (·, ·) denotes the covariance of the two arguments
normalized by the variance of the first argument.

Algorithm Type Input weights w k Output weights g k Performance

GS CS Fixed Global for n bands Adaptive global Good spatial quality with some color

β ( I , MS k ) 1 / n distortions especially on vegetated

areas

GSA-CA CS Adaptive global Adaptive on local window High spatial quality with good color

β ( I , MS k ) Minimization of MSE between P L

and I

preservation.

GLP-CA MRA Not Applicable: details obtained Adaptive on local window High spatial quality with good color

β ( P k , MS k )

by MTF filtering of Pan preservation.

Sensitive to spatial mis-registrations.

GMMSE HYBRID Adaptive global Adaptive global High spectral quality and good spatial

See Equation 10.8 See Equation 10.8 quality.

Notwithstanding D s errors are higher in Table 10.2 than in
Table 10.1, a lower number of blocks is flagged for the IKONOS
test and all the algorithms exhibit a stable behavior, especially on
urban areas.
Finally, Table 10.3 summarizes the main characteristics of the
selected algorithms. Each algorithm is classified according to its
type (CS, MRA and Hybrid) and the input and output weights
are reported for a quick reference. A synthetic judge is also given
as an indication for use.

a performance comparison varying with scale ratios, in Pro-
ceedings of SPIE Image Signal Processing and Remote Sensing V
(ed. S.B. Serpico), 3871, 251-262.
Aiazzi, B., Alparone, L., Argenti, F., Baronti, S. and Pippi, I. (2000)
Multisensor image fusion by frequency spectrum substitution:
subband and multirate approaches for a 3 : 5 scale ratio case,
in Proceedings of the IEEE International Geoscience and Remote
Sensing Symposium , pp. 2629-2631.
Aiazzi, B., Alparone, L., Baronti, S. and Garzelli A. (2002)
Context-driven fusion of high spatial and spectral resolu-
tion data based on oversampled multiresolution analysis.
IEEE Transactions on Geoscience and Remote Sensing , 40 (10),
2300-2312.
Aiazzi,B.,Alparone,L.,Baronti,S.,Garzelli,A.andSelvaM.
(2006) MTF-tailored multiscale fusion of high resolution MS
and Pan imagery. Photogrammetric Engineering &Remote Sens-
ing , 72 (5), 591-596.
Aiazzi, B., Baronti, S. and Selva M. (2007) Improving component
substitution pansharpening through multivariate regression
of MS + Pan data. IEEE Transactions on Geoscience and Remote
Sensing , 45 (10), 3230-3239.
Aiazzi, B., Baronti, S., Lotti, F. and Selva M. (2009) A Com-
parison Between Global and Context Adaptive Pansharpening
of Multispectral Images. IEEE Geoscience and Remote Sensing
Letters , 6 (2), 302-306.
Alparone, L., Cappellini, V., Mortelli, L., Aiazzi, B., Baronti, S.
andCarl a, R. (1998) Apyramid-based approach tomultisensor
image data fusion with preservation of spectral signatures,
in Future Trends in Remote Sensing (ed. P. Gudmandsen)
Balkema, Rotterdam, pp. 418-426.
Alparone, L., Aiazzi, B., Baronti, S. and Garzelli, A. (2003)
Sharpening of very high resolution images with spectral dis-
tortion minimization, in Proceedings of the IEEE International
Geoscience and Remote Sensing Symposium , pp. 21-25.
Alparone, L., Baronti, S., Garzelli, A. and Nencini, F. (2004a)
Landsat ETM

Conclusions

In this chapter the main issues concerning Pan-sharpeningmeth-
ods are introduced and discussed. Some advanced methods are
described in the general frameworks of component substitution
and multiresolution analysis and they are compared with the well
established Gram-Schmidt Pan-sharpening algorithm. Quanti-
tative results evaluated in terms of the quality with no reference
(QNR) index and qualitative results show the efficacy of the
selected algorithms both in terms of spatial and spectral quality
and further improvements appear hard to be obtained. Among
the reviewed algorithms, the context adaptive component sub-
stitution method (GSA-CA) obtains the best scores concerning
QNR, the multiresolution context adaptive algorithm (GLP-CA)
appears the best regarding spatial sharpness in the absence of
misregistration. The hybrid scheme (GMMSE) represents an effi-
cient trade-off since it guarantees good performances with the
advantage of a stable behavior when spatial details are injected.

References

Aiazzi, B., Alparone, L., Baronti, S., Cappellini, V., Carl a, R. and
Mortelli, L. (1997) A Laplacian pyramid with rational scale
factor for multisensor image data fusion, in Proceedings of the
International Conference on Sampling Theory and Application ,
pp. 55-60.
Aiazzi, B., Alparone, L., Argenti, F. and Baronti, S. (1999)
Wavelet and pyramid techniques for multisensor data fusion:

and SAR image fusion based on generalized
intensity modulation. IEEE Transactions on Geoscience and
Remote Sensing , 42 (12), 2832-2839.

+