International Journal of Remote Sensing

ISSN: 0143-1161 (Print) 1366-5901 (Online) Journal homepage: http://www.tandfonline.com/loi/tres20

Review article Multisensor image fusion in remote

sensing: Concepts, methods and applications

C. Pohl & J. L. Van Genderen

To cite this article: C. Pohl & J. L. Van Genderen (1998) Review article Multisensor image fusion
in remote sensing: Concepts, methods and applications, International Journal of Remote Sensing,
19:5, 823-854, DOI: 10.1080/014311698215748

To link to this article: http://dx.doi.org/10.1080/014311698215748

Published online: 25 Nov 2010.

Submit your article to this journal

Article views: 17146

View related articles

Citing articles: 960 View citing articles

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=tres20

Download by: [Xidian University] Date: 08 May 2017, At: 21:20

int. j. remote sensing, 1998 , vol. 19 , no. 5 , 823± 854

Review article

M ultisensor image fusion in remote sensing: concepts, methods

and applications

C. POHL
Western European Union± Satellite Centre (WEUSC), Research Division,
P.O. Box 511, 28850 Torrejo n de Ardoz (Madrid ), Spain

and J. L. VAN GENDEREN

International Institute for Aerospace Survey and Earth Sciences (ITC),
P.O. Box 6, 7500 AA Enschede, The Netherlands

( Received 5 May 1996; in ® nal form 30 May 1997 )

Abstract. With the availability of multisensor, multitemporal, multiresolution

and multifrequency image data from operational Earth observation satellites the
fusion of digital image data has become a valuable tool in remote sensing image
evaluation. Digital image fusion is a relatively new research ® eld at the leading
edge of available technology. It forms a rapidly developing area of research in
remote sensing. This review paper describes and explains mainly pixel based
image fusion of Earth observation satellite data as a contribution to multisensor
integration oriented data processing.

1. Introduction
Earth observation satellites provide data covering di€ erent portions of the electro-
magnetic spectrum at di€ erent spatial, temporal and spectral resolutions. For the
full exploitation of increasingly sophisticated multisource data, advanced analytical
or numerical data fusion techniques are being developed (Shen 1990 ). Fused images
may provide increased interpretation capabilities and more reliable results since data
with di€ erent characteristics are combined. The images vary in spectral, spatial and
temporal resolution and therefore give a more complete view of the observed objects.
It is the aim of image fusion to integrate di€ erent data in order to obtain
more information than can be derived from each of the single sensor data alone
(`1+ 1 = 3’). A good example is the fusion of images acquired by sensors sensitive to
visible/ infrared ( VIR) with data from active synthetic aperture radar (SAR). The
information contained in VIR imagery depends on the multispectral re¯ ectivity of
the target illuminated by sun light. SAR image intensities depend on the charac-
teristics of the illuminated surface target as well as on the signal itself. The fusion of
these disparate data contribute to the understanding of the objects observed.
Image fusion has many aspects to be looked at. Before being able to implement
and use an image fusion approach some of its questions that need to be answered
by the user include:
Ð What is the objective /application of the user?
Ð Which types of data are the most useful for meeting these needs?
0143± 1161/98 $12.00 Ñ 1998 Taylor & Francis Ltd
824 C. Pohl and J. L . van Genderen

Ð Which is the `best’ technique of fusing these data types for that particular
application?
Ð What are the necessary pre-processing steps involved?
Ð Which combination of the data is the most successful?

These and other questions comprise a large number of parameters to be

considered. The most important question to be answered ® rst is: What is the appli-
cation the data is needed for? Knowing that it is possible to de® ne the necessary
spectral and spatial resolution which again has an impact on the selection of the
remote sensing data. The selection of the sensor depends on satellite and sensor
characteristics such as
Ð orbit;
Ð platform;
Ð imaging geometry of optical and radar satellites;
Ð spectral, spatial and temporal resolution ( Pohl and Genderen 1993).
The availability of speci® c data plays an important role too. It depends on the
satellite coverage, operational aspects of the space agency running the satellite,
atmospheric constraints such as cloud cover, ® nancial issues, etc. ( Pohl 1996).
The next step is the choice of a suitable fusion level. The pre-processing steps
are depending on this. In case of pixel based image fusion the geocoding is of vital
importance. Related to the geometric correction of the data, details such as the
Ð geometric model
Ð ground control points (number, distribution, accuracy)
Ð digital elevation model
Ð resampling method etc.

need further consideration.

The question of which technique to choose is very closely related to the de® nition
of evaluation criteria. Both ® elds are rather di cult to de® ne and depend very much
on empirical results (further considerations on this topic are reviewed in § 4 and § 6.2 ).
The application also de® nes which season and weather conditions might be of
relevance to the fused results. Naturally, the same is valid for the observed area.
Especially the topography has a great in¯ uence on the fused remote sensing data
besides the actual ground cover and land use. Another relevant issue in order to
make full use of the bene® ts of image fusion is the selection of appropriate inter-
pretation methods. Particularly when fusing very disparate data sets, e.g., VIR and
SAR the resulting grey values might not refer to physical attributes. The data has
to be carefully evaluated using ground truth for veri® cation ( Pohl 1996, Polidori
and Mangolini 1996 ). Bearing in mind all these considerations further research is
required to generalise and operationalize image fusion.
Image fusion requires well-de® ned techniques as well as a good understanding
of the input data. This review is meant to contribute to the comprehension of image
fusion including the de® nition of terms, the explanation of existing techniques and
the assessment of achievements in image fusion. It is structured in ® ve sections.
Following this introduction a de® nition of image fusion provides the concepts
involved. Then the paper explains why and in which cases image fusion might be
useful. Thereafter, the existing techniques are reviewed including the necessary
processing performances followed by an overview of the evaluation criteria for
 Multisensor image f usion in remote sensing 825

Figure 1. Processing levels of image fusion.

fused data. Finally, bene® ts and limitations of image fusion are summarized in the
concluding section.

2. Concept of image fusion

Data f usion is a process dealing with data and information from multiple sources
to achieve re® ned / improved information for decision making ( Hall 1992). A general
de® nition of image fusion is given as `Image fusion is the combination of two or more
di€ erent images to form a new image by using a certain algorithm’ (Genderen and
Pohl 1994 ).
Image fusion is performed at three di€ erent processing levels according to the
stage at which the fusion takes place:
1. Pixel
2. Feature
3. Decision level.
Image fusion at pixel level means fusion at the lowest processing level referring to
the merging of measured physical parameters. An illustration of the concept of pixel
based fusion is visualised in ® gure 1. It uses raster data that is at least co-registered
but most commonly geocoded. The geocoding plays an essential role because mis-
registration causes arti® cial colours or features in multisensor data sets which falsify
the interpretation later on. It includes the resampling of image data to a common
pixel spacing and map projection, the latter only in the case of geocoding. A
comparison of methods for the geometric and radiometric correction of remote
sensing data is given in Cheng et al. ( 1995 ) and Toutin ( 1994).
Fusion at feature level requires the extraction of objects recognised in the various
data sources, e.g., using segmentation procedures. Features correspond to charac-
teristics extracted from the initial images which are depending on their environment
such as extent, shape and neighbourhood (Mangolini 1994 ). These similar objects
(e.g., regions) from multiple sources are assigned to each other and then fused for
further assessment using statistical approaches or Arti® cial Neural Networks (ANN).
Decision- or interpretation level fusion represents a method that uses value-added
data where the input images are processed individually for information extraction.
The obtained information is then combined applying decision rules to reinforce
826 C. Pohl and J. L . van Genderen

common interpretation and resolve di€ erences and furnish a better understanding
of the observed objects (Shen 1990).
This literature review mainly tackles pixel based image fusion. Techniques used
to fuse data at pixel level are described in § 4. In the literature a large number of
terms can be found. Some terms de® ne di€ erent approaches to image fusion, whilst
others can be used equivalently. Keys et al. ( 1990 ) and Franklin and Blodgett ( 1993 )
describe the fusion as the computation of three new values for a pixel based on the
known relation between the input data for the location in the image. The input
images are di€ erent in terms of spatial, spectral or temporal characteristics. Other
expressions accepted in this context are image merging (Carper et al. 1990), image
integration ( Welch and Ehlers 1988) and multi-sensor data f usion ( Franklin and
Blodgett 1993 ). A broader view of the term is presented by Mangolini ( 1994). He
describes data fusion as group of methods and approaches using multisource data
of di€ erent nature to increase the quality of information contained in the data. Often,
in the case of data fusion, not only remote sensing images are fused but further
ancillary data (e.g., topographic maps, GPS coordinates, geophysical information,
etc.) contribute to the resulting image ( Harris and Murray 1989 ). Data integration
comprises algorithms for image fusion too ( Nandhakumar 1990). This term is often
used in the context of geographical information systems (GIS) ( Ehlers 1993). Other
scientists evaluate multisensor image data in the context of combined ( Lichtenegger
1991 ), coincident (Crist 1984), complementary ( Koopmans and Forero 1993),
composited ( Daily et al. 1979 ) or co-registered ( Rebillard and Nguyen 1982 ) data
analysis. In these cases not always an alteration of the digital numbers amongst the
di€ erent image channels is involved. A simple overlay of multi-source data in a Red-
Green-Blue ( RGB) colour space integrates the data set. The replacement of one of
the three channels or parts of it with an image from another data source is called
substitution (Suits et al. 1988 ). Referring to a di€ erent level of image processing are
the words information f usion (Shufelt and McKeown 1990 ). Here the images are
already interpreted to reach the information or knowledge level before being fused.
A very wide ® eld of applications and approaches in image fusion are summarised by
synergy (Genderen et al. 1994 ) or synergism ( Harris and Graham 1976) of remote
sensing data. It requires the input of data that provide complementary rather than
redundant information.
Apart from the three levels at which fusion can be performed image fusion can
be applied to various types of data sets:
Ð single sensorÐ temporal ( Weydahl 1993 ), e.g., SAR multitemporal for change
detection
Ð multi-sensorÐ temporal ( Pohl and Genderen 1995 ), e.g., VIR/ SAR image
mapping
Ð single sensorÐ spatial (Cliche et al. 1985 ), e.g., high/ low resolution
panchromatic/ multi-spectral SPOT
Ð multi-sensorÐ spatial (Chavez et al. 1991), e.g., high/ low resolution
SPOT/ Landsat
Ð single dataÐ multi-sensor (Guyenne 1995 ), e.g., ERS-1/ ERS-2
Ð remote sensing data with ancillary data ( Janssen et al. 1990), e.g., image with
topographic maps, (the references given are not exhaustive but meant as
example).
Having introduced the concept of image fusion, the following section outlines some
goals of image fusion.
 Multisensor image f usion in remote sensing 827

3. Objectives of image fusion

Image fusion is a tool to combine multisource imagery using advanced image
processing techniques. It aims at the integration of disparate and complementary
data to enhance the information apparent in the images as well as to increase the
reliability of the interpretation. This leads to more accurate data ( Keys et al. 1990)
and increased utility ( Rogers and Wood 1990). It is also stated that fused data
provides for robust operational performance, i.e., increased con® dence, reduced
ambiguity, improved reliability and improved classi® cation ( Rogers and Wood 1990).
Studying the literature, it appears that image fusion is applied to digital imagery in
order to:
Ð sharpen images (Chaves et al. 1991);
Ð improve geometric corrections (Strobl et al. 1990)
Ð provide stereo-viewing capabilities for stereophotogrammetry ( Bloom et al.
1988 )
Ð enhance certain features not visible in either of the single data alone ( Leckie
1990 )
Ð complement data sets for improved classi® cation (Schistad-Solberg et al. 1994)
Ð detect changes using multitemporal data ( Duguay et al. 1994)
Ð substitute missing information (e.g., clouds-VIR, shadows-SAR) in one image
with signals from another sensor image (Aschbacher and Lichtenegger 1990)
Ð replace defective data ( Suits et al. 1988).
The following paragraphs describe the achievements of image fusion in more detail.

3.1. Image sharpening

Image fusion can be used as a tool to increase the spatial resolution. In that
case high-resolution panchromatic imagery is fused with low-resolution often multi-
spectral image data. A distinction has to be made between the pure visual enhance-
ment (superimposition) and real interpolation of data to achieve higher resolution
(wavelets) the latter being proposed amongst others by Mangolini ( 1994 ) and Ranchin
et al. ( 1996 ). In this way the spectral resolution may be preserved while a higher
spatial resolution is incorporated which represents the information content of the
images in much more detail ( Franklin and Blodgett 1993, Pellemans et al. 1993 ).
Examples of fusing XS/ PAN or TM / PAN for resolution enhancement were published,
amongst others, by Simard ( 1982, simulated data) Cliche et al. ( 1985), Price ( 1987),
Carper et al. ( 1990 ), Franklin and Blodgett ( 1993), Ranchin et al. ( 1996 ). A special
case forms the fusion of channels from a single sensor for resolution enhancement,
e.g., TM-data. The lower resolution thermal channel can be enhanced using the
higher resolution spectral channels (Moran 1990). Other approaches increase the
spatial resolution of the output channel using a windowing technique on the six
multispectral bands of TM (Sharpe and Kerr 1991). The fusion of SAR / VIR does
not only result in the combination of disparate data but may also be used to spatially
enhance the imagery involved. Geometric accuracy and increase of scales using
fusion techniques is of concern to mapping and map updating (Chiesa and Tyler
1990, Pohl 1996 ).

3.2. Improvement of registration accuracy

The multi-sensor image registration gains importance in areas with frequent
cloud cover where VIR imagery is not continuously available. The drawbacks of
828 C. Pohl and J. L . van Genderen

conventional image to map registration are the di€ erences in appearance of the
features used as control or tie points. The positioning of points is facilitated if they
are localised in similar `views’ (remote sensing data). Of advantage is an integrative
recti® cation approach which iteratively improves the registration accuracy ( Ehlers
1991 ). Multisensor stereo mapping requires accurately co-registered input data and
shows improved results in comparison with image-to-map registration ( Welch et al.
1985, 1990 ).

3.3. Creation of stereo data sets

It is proven that multisensor stereo data sets can help to overcome the lack of
information, e.g., due to cloud cover. Combinations of VIR/ VIR (di € erent spatial
resolution), SAR / SAR (multiple incidence angles) and VIR/ SAR were successfully
used. Some constraints exist depending on radiometric di€ erences of the images used
as stereo pair ( Domik et al. 1986, Bloom et al. 1988, Welch et al. 1990, Buchroithner
et al. 1992, Raggam et al. 1994, Toutin and Rivard 1995 ).

3.4. Feature enhancement

Taking advantage of the di€ erent physical nature of microwave and optical sensor
systems the fusion of those data results in an enhancement of various features
observed ( Daily et al. 1979, Ehlers 1991, Franklin and Blodgett 1993, Ye sou et al.
1993 b). The feature enhancement capability of image fusion is visually apparent in
VIR/ VIR combinations that often results in images that are superior to the original
data ( Keys et al. 1990 ). Multisensor image fusion enhances semantic capabilities of
the images and yields information which is otherwise unavailable or hard to obtain
from single sensor data (Mitiche and Aggarwal 1986 ). Conclusively, in order to
maximise the amount of information extracted from satellite image data useful
products can be found in fused images ( Welch and Ehlers 1987).

3.5. Improved classi® cation

The classi® cation accuracy of remote sensing images is improved when multiple
source image data are introduced to the processing. Images from microwave and
optical sensors o € er complementary information that helps in discriminating the
di€ erent classes. Working with VIR data users rely on the spectral signature of the
ground targets in the image. Some vegetation species can not be separated due to
their similar spectral response. Therefore radar images can contribute di€ erent signals
due to di€ erences in surface roughness, shape and moisture content of the observed
ground cover. The use of multisensor data in image classi® cation becomes more
and more popular with the increased availability of sophisticated software and
hardware facilities to handle the increasing volumes of data. A new trend in this
respect is the ANN approach since the conventional statistical techniques were found
inappropriate for processing multisensor data from SAR and VIR.

3.6. T emporal aspects for change detection

Image fusion for change detection takes advantage of the di€ erent con® gurations
of the platforms carrying the sensors. Based on the orbital characteristics the so
called repeat-cycle of the satellite is de® ned and varies from system to system from
daily ( NOAA) up to tenths of days ( Landsat, SPOT, ERS-1, etc.). This means that
a certain area on the Earth is observed at di€ erent times of the day, month or year.
The combination of these temporal images enhances information on changes that
 Multisensor image f usion in remote sensing 829

might have occurred in the area observed. T emporal image f usion is applicable to
images from the same sensor as well as to multiple sensor data. It is nearly impossible
to acquire multisensor data simultaneously so that fusion of data from di€ erent
sensors mostly includes a temporal factor. An overview on image enhancement
techniques for change detection from multisensor data including image di€ erencing,
ratioing and PCA is given in Mouat et al. ( 1993). The techniques are described in
§ 4. There is a necessity of correcting the input imagery for radiometric distortions
which occur especially in VIR data, i.e., atmospheric and illumination correction in
order to create compatible data sets. The images to be fused should be acquired at
similar seasons to account for seasonal changes which might in¯ uence the change
detection capabilities of this approach.
Multitemporal SAR data provides good potential for change detection analysis
because of its all-weather capability. The evaluation of single SAR imagery is di cult
due to the ambiguities contained in SAR images ( De Groof et al. 1992, Weydahl
1992, Kattenborn et al. 1993, Kohl et al. 1994 ).

3.7. Overcoming gaps

The images acquired by satellite remote sensing are in¯ uenced by a number of
e€ ects based on the carrier frequency of the electromagnetic waves. It is well known
that VIR sensors are hindered by clouds to obtain information on the observed
objects on Earth. Even the shadows of the clouds in¯ uence the interpretability of
the imagery. SAR on the other hand su€ ers from severe terrain induced geometric
distortions based on its side-looking geometry ( layover, foreshortening, shadow). To
overcome these in¯ uences it is possible to combine di€ erent images acquired by the
same or a di€ erent instrument. Apart from creating simple or complex mosaics there
are other techniques such as Optimal Resolution Approach ( Haefner et al. 1993 ),
IHS and others to fuse the data for the purpose of ® lling the gaps which are described
in more detail in the following section.

4. Image fusion techniques

This section describes techniques to fuse remote sensing images. The evaluation
of their usefulness and disadvantages is discussed later in § 6. An overall processing
¯ ow for fusion optical and radar satellite data is given in ® gure 2.
After having corrected the remote sensing images for system errors, the data is
further radiometrically processed. In the case of SAR, speckle reduction is an
elementary operation in many applications. The speckle reduction can be performed
at various stages of the processing; depending on the application, it is advisable to
speckle ® lter before geocoding which implies an improved object identi® cation for
GCP measurements ( Dallemand et al. 1992 ), others allow the ® ltering and resampling
during the geocoding process in one step to reduce the number of times the data is
resampled. Another aspect to be considered when dealing with SAR is the 16 to
8-bit data conversion. According to Knipp ( 1993) the 16 to 8-bit data conversion in
SAR processing, in case that it is required, should be performed after speckle
reduction in order to reduce the loss of information. Optical imagery is in¯ uenced
by the atmosphere during data acquisition and therefore needs correction and /or
other radiometric enhancements such as edge enhancement and others.
Following the radiometric processing the data are geometrically corrected. In
some cases it is geocoded, in others only co-registered to coincide on a pixel by pixel
basis depending on the height variations in the area contained in the data. Again it
830 C. Pohl and J. L . van Genderen

Figure 2. Processing ¯ ow chart for pixel based image fusion ( Pohl 1996 ).

has to be pointed out that this is a major element of pixel level image fusion. These
fusion techniques are very sensitive to misregistration.
Subsequently, the data can be fused using one of the fusion techniques described
in this review. In some cases, especially if image data of very di€ erent spatial
resolution is involved, the resampling of the low resolution data to the pixel size of
the high resolution image might causes a blocky appearance of the data. Therefore
a smoothing ® lter can be applied before actually fusing the images (Chavez 1987).
The resulting image map can be further evaluated and interpreted related to the
desired application.
In general, the techniques can be grouped into two classes: ( 1 ) Colour related
techniques, and ( 2 ) Statistical/ numerical methods. The ® rst comprises the colour
composition of three image channels in the RGB colour space as well as more
sophisticated colour transformations, e.g., IHS and HSV. Statistical approaches are
developed on the basis of channel statistics including correlation and ® lters.
Techniques like PCA and regression belong to this group. The numerical methods
follow arithmetic operations such as image di€ erencing and ratios but also adding
of a channel to other image bands. A sophisticated numerical approach uses wavelets
in a multiresolution environment. The next sections describe these techniques in
more detail.

4.1. Band selection

Some techniques only allow a limited number of input bands to be fused
(e.g., RGB, IHS), whilst others can be performed with any number of selected input
bands. One method that relies on statistics in order to select the data containing
most of the variance. This is the selection method developed by Chavez et al. ( 1982)
called Optimum Index Factor (OIF) mathematically described in equation ( 1).
3
ž si
OIF = =
i 1
3 ( 1)
ž |ccj |
j 1
=
s i = standard deviation of digital numbers for band, cc j = correlation coe cient
 Multisensor image f usion in remote sensing 831

between any two of three bands. Other considerations related to OIF can be found
in Chavez et al. ( 1984 ), Keys et al. ( 1990 ) and Ye sou et al. ( 1993 a).
Another approach is to select the bands which are the most suitable to a certain
application. This requires a priori knowledge by the user (She eld 1985 ). Kaufmann
and Buchroithner ( 1994) suggest to select the three bands with the highest variance.
The principal component analysis is another solution to reduce the number of
channels containing the majority of the variance for the purpose of image fusion
( Ye sou et al. 1993 a). Others provide a certain band combination based on an
algorithm which takes into account the statistics of the scene including correlations
between image channels (She eld 1985 ).

4.2. Colour related techniques

There are a variety of techniques to display image data in colour. In general,
colours are mathematically described in two ways:
1. Tristimulus values,
2. Chromaticity
Each of these has several special cases. Tristimulus values are based on three di€ erent
spectral curves. They can represent ® lters or emission spectra. A special representation
of the spectral curves are the CIE tristimulus values obtained with the CIE tristimulus
coe cients (see ® gure 3).
A chromaticity representation consists of the luminosity of an object and two
quotients of the luminosity leading to intensity hue and saturation. A chromatic
representation based on the CIE tristimulus values and using Cartesian coordinates
is shown in ® gure 3. RGB refers to the tristimulus values associated with a
colour monitor. Its advantages are its simplicity and the fact that other colour
representations have to be transformed to RGB in order to be displayed on a colour
monitor ( Haydn et al. 1982, Russ 1995). It is one of the most common techniques
to colour composite multi-sensor data and is mathematically described ® rst in the
following section.
4.2.1. Colour composites (RGB)
The so-called additive primary colours allow one to assign three di€ erent types
of information (e.g., image channels) to the three primary colours red, green and

Figure 3. CIE chromaticity diagram (Haydn et al. 1982).

832 C. Pohl and J. L . van Genderen

blue. Together they form a colour composite that can be displayed with conventional
media, e.g., cathode ray tube (CRT). The colour composite facilitates the interpreta-
tion of multichannel image data due to the variations in colours based on the values
in the single channels. The grey scale value, used to select a set of red, green, and
blue brightness’, is stored in a look up table ( LUT ) that are the voltages sent to the
display tube. Operations on the LUT and the histogram of the image data can
enhance the colour composite for visual interpretation ( Russ 1995 ).
The possibilities of varying the composite are manifold. Depending on the selec-
tion of the input image channels the fused data will show di€ erent features. Very
important for the colour composite is the distribution of the available 0± 255 grey
values to the range of the data. It might be of advantage to invert input channels
before combining them in the RGB display with other data depending on the objects
of interest to be highlighted ( Pohl et al. 1994 ).
Examples of successfully used colour composites of optical and microwave data
are described by Aschbacher and Lichtenegger ( 1990), Dallemand et al. ( 1992),
Vornberger and Bindschadler ( 1992), Comhaire et al. ( 1994), Hinse and Coulombe
( 1994 ), Marek and Schmidt ( 1994), Oprescu et al. ( 1994), and Pohl et al. ( 1994 ).
Reports about multi-sensor optical composites can be found in Welch et al. ( 1985 ),
and Chavez ( 1987 ). Multisensor SAR fusion by RGB are reported by Marek and
Schmidt ( 1994 ). Multitemporal ERS-1 SAR colour composites were used by
Comhaire et al. ( 1994). On the basis of the complementary information from VIR
(spectral re¯ ectivity) and SAR (surface roughness) the features are enhanced in fused
imagery. Work in the ® eld of geology was published amongst others by Daily et al.
( 1979 ), Zobrist et al. ( 1979) and Ye sou et al. ( 1993 b).
In many cases the RGB technique is applied in combination with another image
fusion procedure, e.g., IHS, PCA, and others which are explained in the coming
sections.

4.2.2. Intensity-Hue-Saturation (IHS)

The IHS colour transformation e€ ectively separates spatial (I ) and spectral (H, S )
information from a standard RGB image. It relates to the human colour perception
parameters. The mathematical context is expressed by equation ( 2) (a± c). I relates
to the intensity, while `n 1’ and `n 2’ represent intermediate variables which are needed
in the transformation. H and S stand for Hue and Saturation, ( Harrison and

A B
Jupp 1990 ).

1 1 1

AB AB
Ó 3 Ó 3 Ó 3
I R
1 1 2
n1 = Õ G (a)
Ó 6 Ó 6 Ó 6
n2 B ( 2)
1 1
Õ 0
Ó 2 Ó 2

AB
n2
H = tan Õ S = Ó n1 + n2
1 2 2
(b) (c)
n1

There are two ways of applying the IHS technique in image fusion: direct and
substitutional. The ® rst refers to the transformation of three image channels assigned
 Multisensor image f usion in remote sensing 833

to I , H and S ( Rast et al. 1991). The second transforms three channels of the data
set representing RGB into the IHS colour space which separates the colour aspects
in its average brightness (intensity). This corresponds to the surface roughness, its
dominant wavelength contribution ( hue) and its purity (saturation) (Gillespie et al.
1986, Carper et al. 1990). Both the hue and the saturation in this case are related to
the surface re¯ ectivity or composition (Grasso 1993). Then, one of the components
is replaced by a fourth image channel which is to be integrated. In many published
studies the channel that replaced one of the IHS components is contrast stretched
to match the latter. A reverse transformation from IHS to RGB as presented in
equation ( 3 ) converts the data into its original image space to obtain the fused image
( Hinse and Proulx 1995 b).
The IHS technique has become a standard procedure in image analysis. It serves
colour enhancement of highly correlated data (Gillespie et al. 1986), feature enhance-
ment (Daily 1983), the improvement of spatial resolution (Welch and Ehlers 1987,

A B
Carper et al. 1990 ) and the fusion of disparate data sets ( Harris et al. 1990, Ehlers 1991).

1 1 1

AB AB
Ó 3 Ó 6 Ó 2
R I
1 1 1
G = Õ n1 ( 3)
Ó 3 Ó 6 Ó 2
B n2
1 2
Õ 0
Ó 3 Ó 6

The use of IHS technique in image fusion is manifold, but based on one principle:
the replacement of one of the three components (I , H or S ) of one data set with
another image. Most commonly the intensity channel is substituted. Replacing the
intensityÐ the sum of the bandsÐ by a higher spatial resolution value and reversing
the IHS tranformation leads to composite bands (Chavez et al. 1991 ). These are
linear combinations of the original (resampled ) multispectral bands and the higher
resolution panchromatic band (Campbell 1993).
A variation of the IHS fusion method applies a stretch to the hue saturation
components before they are combined and transformed back to RGB ( Zobrist et al.
1979 ). This is called colour contrast stretching (Gillepsie et al. 1986). The IHS
transformation can be performed either in one or in two steps. The two step approach
includes the possibility of contrast stretching the individual I , H and S channels. It
has the advantage of resulting in colour enhanced fused imagery ( Ehlers 1991). More
results using IHS image fusion are reported by Rast et al. ( 1991 ), Jutz and Chorowicz
( 1993 ), Koopmans and Richetti ( 1993), Oprescu et al. ( 1994), Smara et al. ( 1996 ),
and Yildimi et al. ( 1996 ). A closely related colour system to IHS (sometimes also
called HSI) is the HSV: hue, saturation and value ( Russ 1995 ). An example of HSV
image fusion was presented by Chiesa and Tyler ( 1990).

4.2.3. L uminance-chrominance
Another colour encoding system called YIQ has a straightforward transformation
from RGB with no loss of information. Y , the luminance , is just the brightness of a
panchromatic monochrome image. It combines the red, green, and blue signals in
proportion to the human eye’s sensitivity to them. The I and Q components of the
colour are chosen for compatibility with the hardware used. The I is essentially red
834 C. Pohl and J. L . van Genderen

minus cyan, while Q is magenta minus green. The relation between YIQ and RGB
is shown in equation ( 4 a , b ):

AB A BA B
Y 0´299 0´587 0´114 R

(a) I = 0´596 Õ 0´274 Õ 0´322 G

Q 0´211 Õ 0´523 0´312 B

AB A BA B
( 4)
R 1´000 0´956 0´621 Y

(b) G = 1´000 Õ 0´272 Õ 0´647 I

B 1´000 Õ 1´106 Õ 1´703 Q

Since the Y , I and Q components are less correlated than the RGB ones, this
transformation o € ers better possibilities for enhancing an image. This was shown in
an example by Guo and Pinliang ( 1989).

4.3. Statistical / numerical methods

In this section all operations are grouped together which deal with mathematical
combinations of image channels. It comprises addition, subtraction, multiplication
and division (ratio) as well as combinations of these operations.
4.3.1. Arithmetic combinations
Certain combinations of bands of images can lead to image sharpening, i.e.,
higher spatial resolution. Common combinations are SPOT PAN and SPOT XS or
SPOT PAN and Landsat TM. But the integration of SAR can also improve the
spatial structure of an image because it introduces the surface roughness to the image.
4.3.1.1. Adding and multiplicati on
To enhance the contrast, adding and multiplication of images are useful. An
example is the multiplication process expressed by equation ( 5 a , b ):
DNf = A (w 1 DNa + w 2 DNb ) + B ( 5 a)
DNf = A DNa DNb + B ( 5 b)
A and B are scaling factors and w 1 and w 2 weighting parameters. DNf , DN a and
DNb refer to digital numbers of the ® nal fused image and the input images a and b ,
respectively. This method was successfully applied to Landsat-TM and SPOT PAN
data ( Ye sou et al. 1993 b).
There are a large number of publications containing suggestions on how to fuse
high resolution panchromatic images with lower resolution multi-spectral data to
obtain high resolution multi-spectral imagery. The possibilities of combining the
data using multiplication or summation are manifold. The choice of weighting and
scaling factors may improve the resulting images. Details can be found in Simard
( 1982 ), Cliche et al. ( 1985 ), Pradines ( 1986), Price ( 1987), Welch and Ehlers ( 1987 ),
Carper et al. ( 1990 ), Ehlers ( 1991 ), Mangolini et al. ( 1993), Munechika et al. ( 1993 )
and Pellemans et al. ( 1993).
4.3.1.2. Di€ erence and ratio images
Di € erence or ratio images are very suitable for change detection (Mouat et al.
1993 ). The ratio method is even more useful because of its capability to emphasise
more on the slight signature variations ( Zobrist et al. 1979, Singh 1989). In some
 Multisensor image f usion in remote sensing 835

cases the resulting di€ erence image contains negative values. Therefore a constant
has to be added to produce positive digital numbers. However, di€ erences do not
always refer to changes since other factors, like di€ erences in illumination, atmo-
spheric conditions, sensor calibration, ground moisture conditions and registration
of the two images, can lead to di€ erences in radiance values. In ratioing, two images
from di€ erent dates are divided, band by band if the image data have more than
one channel. If the intensity of the re¯ ected energy is nearly the same in each image
then the ratio image pixel is one, it indicates no change. The critical part of this
method is selecting appropriate threshold values in the lower and upper tails of the
distribution representing change pixel values. In this respect the normalisation of
the data is of advantage as indicated in equation ( 6 ), (Gri ths 1988):
XS3 Õ XS 2 TM 4 Õ TM 3
Õ +C (C = 128 for positive values) ( 6)
XS3 + XS2 TM 4 + TM 3
A ratio for spatial enhancement is summarised by equation ( 7). The aim of this
method is to maintain the radiometric integrity of the data while increasing the
spatial resolution.
DNXS(i)
DNHybridXS(j) = DNPAN ( 7)
DNSynPAN
where DNHybridXS(i) - i th band of the fused high resolution image, DNPAN - correspond-
ing pixel in high resolution input PAN image, DNXS(i) - super-pixel in i th band of
input low resolution XS image, DNSynPAN - corresponding pixel in low resolution
synthetic PAN image, created from low resolution multispectral bands that overlap
the spectral response of the input high resolution PAN, (Munechika et al. 1993 ).
That is also the aim of the Brovey Transform, named after its author. It is a
formula that normalises multispectral bands used for a RGB display, and multiplies
the result by any other desired data to add the intensity or brightness component
to the image. The algorithm is shown in equation ( 8) where D N fused means the DN
of the resulting fused image produced from the input data in n multispectral bands
multiplied by the high resolution image DNhighres .
DNb1
DNfused = DNhighres ( 8)
DNb1 + DNb2 + DNbn
4.3.2. Principal component analysis
The PCA is useful for image encoding, image data compression, image enhance-
ment, digital change detection, multitemporal dimensionality and image fusion. It is
a statistical technique that transforms a multivariate data set of intercorrelated
variables into a data set of new un-correlated linear combinations of the original
variables. It generates a new set of axes which are orthogonal.
The approach for the computation of the principal components ( PCs) comprises
the calculation of:
1. Covariance (unstandardised PCA) or correlation (standardised PCA) matrix
2. Eigenvalues, -vectors
3. PCs
An inverse PCA transforms the combined data back to the original image space.
The use of the correlation matrix implies a scaling of the axes so that the features
836 C. Pohl and J. L . van Genderen

receive a unit variance. It prevents certain features from dominating the image
because of their large digital numbers. The signal-to-noise ratio ( SNR) is signi® cantly
improved applying the standardised PCA (Singh and Harrison 1985, Shettigara
1992 ). Better results are obtained if the statistics are derived from the whole study
area rather than from a subset area ( Fung and LeDrew 1987). The PCA technique
can also be found under the expression Karhunen L oeve approach ( Zobrist et al.
1979 ).
Two types of PCA can be performed: selective or standard. The latter uses all
available bands of the input image, e.g., TM 1± 7, the selective PCA uses only a
selection of bands which are chosen based on a priori knowledge or application
purposes ( Ye sou et al. 1993 b). In case of TM the ® rst three PCs contain 98± 99 per
cent of the variance and therefore are su cient to represent the information.
PCA in image fusion has two approaches:
1. PCA of multichannel image Ð replacement of ® rst principal component by
di€ erent images (Principal Component Substitution Ð PCS) (Chavez et al.
1991 ) or
2. PCA of all multi-image data channels ( Ye sou et al. 1993 a).
The ® rst version follows the idea of increasing the spatial resolution of a multichannel
image by introducing an image with a higher resolution. The channel which will
replace PC1 is stretched to the variance and average of PC1. The higher resolution
image replaces PC1 since it contains the information which is common to all bands
while the spectral information is unique for each band (Chavez et al. 1991); PC1
accounts for maximum variance which can maximise the e€ ect of the high resolution
data in the fused image (Shettigara 1992 ).
The second procedure integrates the disparate natures of multisensor imput data
in one image. The image channels of the di€ erent sensor are combined into one
image ® le and a PCA is calculated from all the channels. Some examples of image
fusion applying the ® rst and the second method of PCA are reported by Ye sou et al.
( 1993 a) and Richards ( 1984 ), respectively.
A similar approach to the PCS is accomplished in the C -stretch (colour stretch)
( Rothery and Francis 1987) and the D -stretch (de-correlation stretch) ( Ehlers 1987,
Campbell 1993, Jutz and Chorowicz 1993 ). The de-correlation stretch helps to
overcome the perceived problem that the original data often occupy a relatively
small portion of the overall data space (Campbell 1993 ). In D -stretching three
channel multispectral data are transformed on to principal component axes, stretched
to give the data a spherical distribution in feature space and then transformed back
onto the original axes ( Jutz and Chorowicz 1993). In C -stretching PC1 is discarded,
or set to a uniform DN across the entire image, before applying the inverse trans-
formation. This yields three colour stretched bands which, when composited, retain
the colour relations of the original colour composite but albedo and topographically
induced brightness variations are removed.
The PCA approach is sensitive to the choice of area to be analysed. The
correlation coe cient re¯ ects the tightness of a relation for a homogeneous sample.
However, shifts in the band values due to markedly di€ erent cover types also
in¯ uence the correlations and particularly the variances (Campbell 1993).
4.3.3. High pass ® ltering
Another approach to enhance the spatial resolution of multispectral data adds
the spatial to the spectral information: high pass ® ltering ( HPF) in combination
 Multisensor image f usion in remote sensing 837

with band addition. The high spatial resolution image is ® ltered with a small high-
pass-® lter resulting in the high frequency part of the data which is related to the
spatial information. This is pixel wise added to the low resolution bands ( Tauch and
KaÈhler 1988, Shettigara 1992, Jutz and Chorowicz 1993). Nevertheless, the HPF
method has limitations in passing on important textural information from the high
resolution band to the low resolution data (Shettigara 1992).

4.3.4. Regression variable substitution

Multiple regression derives a variable, as a linear function of multi-variable
data, that will have maximum correlation with unvariate data. In image fusion the
regression procedure is used to determine a linear combination (replacement vector)
of an image channel that can be replaced by another image channel. This method
is called regression variable substitution ( RVS).
To achieve the e€ ect of fusion the replacement vector should account for a
signi® cant amount of variance or information in the original multi-variate data set.
The method can be applied to spatially enhance data or for change detection with
the assumption that pixels acquired at time one are a linear function of another set
of pixels received at time two. Using the predicted value obtained from the least-
square regression, the di€ erence image is the regression value-pixel of time one (Singh
1989, Shettigara 1992 ).

4.3.5. Canonical variate substitution

The canonical variate analysis is a technique which produces new composite
bands based on linear combinations of the original bands. It derives linear
combinations of the bands which maximise the di€ erences between training or
reference classes, relative to the variation within the training classes. Successful
implementation of this approach requires the de® nition of homogeneous areas or
training classes for each of the information classes of interest (Campbell 1993 ).

4.3.6. Component substitution

Shettigara ( 1992 ) presented a slightly di€ erent view of categorising certain image
fusion techniques. He suggests to look at it in a more general way and calls any
technique that involves a forward transformation, the replacement of a variable and
a backward transformation, a COmponent Substitution (COS) technique. The aim of
this technique is to sharpen imagery. The most popular COS technique (already
described above) is the IHS colour tranformation. Other methods which can be ® tted
into the COS model are Regression V ariable Substitution ( RVS), Principal Component
Substitution ( PCS) and Standardised PC Substitution (SPS), already discussed in the
preceding sections (Shettigara 1992).

4.3.7. W avelets
A mathematical tool developed originally in the ® eld of signal processing can
also be applied to fuse image data following the concept of the multiresolution analysis
(MRA) (Mallat 1989). Another application is the automatic geometric registration
of images, one of the pre-requisites to pixel based image fusion ( Djamdji et al. 1993).
The wavelet transform creates a summation of elementary functions (= wavelets)
from arbitrary functions of ® nite energy. The weights assigned to the wavelets are
the wavelet coe cients which play an important role in the determination of structure
characteristics at a certain scale in a certain location. The interpretation of structures
838 C. Pohl and J. L . van Genderen

or image details depend on the image scale which is hierarchically compiled in a

pyramid produced during the MRA ( Ranchin and Wald 1993).
The wavelet transform in the context of image fusion is used to describe di€ erences
between successive images provided by the MRA. Once the wavelet coe cients are
determined for the two images of di€ erent spatial resolution, a transformation model
can be derived to determine the missing wavelet coe cients of the lower resolution
image. Using these it is possible to create a synthetic image from the lower resolution
image at the higher spatial resolution. This image contains the preserved spectral
information with the higher resolution, hence showing more spatial detail. This
method is called ARSIS, an abbreviation of the French de® nition `ame lioration de
la re solution spatial par injection de structures’ ( Ranchin et al. 1996 ).

4.4. Combined approaches

In some cases it might not be enough to follow one approach in order to achieve
the required results. Therefore, the use of combined techniques plays an important
role in image fusion. The most successful techniques for using images are developed
from combinations of techniques, such as multiple band combinations, mosaic and
other fusion techniques. Aspects such as:
(a ) cloud cover problem not solved,
(b) blurred image,
(c) colour assignment not acceptable,
(d ) data stretch insu cient,
are reasons for investigating other possibilities and the optimization of techniques.
The following paragraphs provide a brief overview on the manifold possibilities of
combining techniques in order to fuse remote sensing images. They serve only as
examples of the combination of techniques used by various authors. Many other
combinations are of course also possible.
4.4.1. RGB/ other techniques
The display of remote sensing data in RGB can be found with many of the fusion
techniques mentioned above. The combinations of IHS / RGB and PCA / RGB found
recognition especially in geology ( Haydn et al. 1982, Singh 1989, Koopmans and
Richetti 1993, Ye sou et al. 1993 a, 1993 b). The arithmetic combination of image
channels in combination with RGB display facilitate the interpretation in many
applications (Lichtenegger et al. 1991 ).
4.4.2. IHS / other techniques
The IHS transformation o € ers a large number of possibilities to fuse remote
sensing data with e.g., ratios and PCA as input. It is widely used to integrate
especially very disparate data such as VIR and SIR ( Ehlers 1991, Chiuderi and Fini
1993, Grasso 1993, Loercher and Wever 1994). Arithmetic methods combined with
IHS including weighting factors can enhance the remote sensing data for visual
interpretation ( Ye sou et al. 1993 a).
4.4.3. HPF/Band combinations
It can be of advantage to use subtraction of pre-processed data (e.g., HPF ® ltered
imagery) from the original data in order to enhance lines and edges. The resulting
data can then be fused using multiplication with SAR imagery. This creates sharpened
VIR/ SAR image combinations. A further development of the method described is
the introduction of multispectral information to the merge of high-resolution PAN
 Multisensor image f usion in remote sensing 839

with SAR. The results of the triple sensor image fusion combine VIR/ SAR and high
spatial resolution in one image ( Pohl 1996).
4.4.4. Mosaic / other techniques
In order to solve the cloud cover problem e€ ectively, the mosaic approach o € ers
a wide variety of possibilities in connection with other fusion techniques. SAR data
is introduced to areas of no information on the optical data, i.e., clouds and their
shadows. Likewise, SAR data from di€ erent sensors or orbits can reduce the regions
of foreshortening, layover and shadow. The idea to input optical data often fails
because the mountainous areas which are causing these geometric distortions in the
SAR are also the reason for cloud coverage. Therefore, the contribution of other
SAR imagery represents the more operational solution. Once the cloud / shadow
( VIR) and layover/ foreshortening / shadow mask (SAR) has been produced, it is
possible to introduce all types of data in the various elements of the mask ( Pohl 1996 ).

5. Ex amples of image fusion

There is an increasing number of applications in which multisensor images are
used to improve and enhance image interpretation. This section gives a couple of
examples of multisensor image fusion comprising the combination of multiple images
and ancillary data with remote sensing images:
Ð topographic mapping and map updating,
Ð land use, agriculture and forestry,
Ð ¯ ood monitoring,
Ð ice- and snow-monitoring and
Ð geology.
Each section contains a list of references for further reading on these topics.

5.1. T opographic mapping and map updating

Image fusion as tool for topographic mapping and map updating has its
importance in the provision of up-to-date information. Areas that are not covered
by one sensor might be contained in another. In the ® eld of topographic mapping
or map updating often combinations of VIR and SAR are used. The optical data
serves as reference whilst the SAR data that can be acquired at any time provides
the most recent situation. In addition the two data sets complement each other in
terms of information contained in the imagery. Work in this ® eld has been published
amongst others in ( Essadiki 1987, Bloom et al. 1988, Tauch and KaÈhler 1988, Welch
and Ehlers 1988, Tanaka et al. 1989, Rogers and Wood 1990, Dallemand et al. 1992,
Perlant 1992, Albertz and Tauch 1994, Kaufmann and Buchroithner 1994, Matte
1994, Perlant et al. 1994, Hinse and Proulx 1995 a, Pohl 1995, 1996, Pohl and
Genderen 1995 ).

5.2. L and use, agriculture and forestry

Regarding the classi® cation of land use the combination of VIR with SAR data
helps to discriminate classes which are not distinguishable in the optical data alone
based on the complementary information provided by the two data sets ( Toll 1985,
Munechika et al. 1993, Franklin and Blodgett 1993, Hussin and Shaker 1996).
Similarly, crop classi® cation in agriculture applications is facilitated (Ahern et al.
1978, Ulaby et al. 1982, Brisco and Brown 1995). Concerning multisensor SAR
image fusion the di€ erence in incidence angles data may solve ambiguities in the
840 C. Pohl and J. L . van Genderen

classi® cation results ( Brisco et al. 1983). Multitemporal SAR is a valuable data
source in countries with frequent cloud cover and successfully used in crop
monitoring. Especially, for Developing Countries the fusion of SAR data with VIR
is a cost e€ ective approach which enables continuous monitoring ( Nezry et al. 1993,
Mangolini and Arino 1996 a, 1996 b). Optical and microwave image fusion is also
well known for the purpose of identifying and mapping forest cover and other types.
The combined optical and microwave data provide a unique combination that allows
more accurate identi® cation, as compared to the results obtained with the individual
sensors ( Leckie 1990, Lozano-Garcia and Ho€ er 1993, Kachwalha 1993, Hinse
and Coulombe 1994, Aschbacher et al. 1994, Wilkinson et al. 1995). With the
implementation of fusion techniques using multisensor optical data the accuracy of
urban area classi® cation is improved mainly due to the integration of multispectral
with high spatial resolution (Gri ths 1988, Haack and Slonecker 1991, Ranchin
et al. 1996 ).

5.3. Flood monitoring

In the ® eld of the management of natural hazards and ¯ ood monitoring using
multisensor VIR/ SAR images plays an important role. In general there are two
advantages to introduce SAR data in the fusion process with optical imagery:
1. SAR is sensitive to the di-electric constant which is an indicator for the
humidity of the soil. In addition, many SAR systems provide images in which
water can be clearly distinguished from land.
2. SAR data is available at any time of the day or year independent from cloud
cover or daylight. This makes it a valuable data source in the context of
regular temporal data acquisition necessary for monitoring purposes.
For the representation of the pre-¯ ood situation the optical data provides a good
basis. The VIR image represents the land use and the water bodies before ¯ ooding.
Then, SAR data acquisition at the time of the ¯ ood can be used to identify ¯ ood
extent and damage. Examples of multisensor fusion for ¯ ood monitoring are
described by Corves ( 1994), Ye sou et al. ( 1994), Pohl et al. ( 1995), Wang et al. ( 1995)
and Fellah and Tholey ( 1996 ). Others rely on multitemporal SAR image fusion to
assess ¯ ood extents and damage ( Matthews and Ga€ ney 1994, ESA 1995 a, 1995 b,
Lichtenegger and Calabresi 1995, MacIntosh and Profeti 1995, Badji 1995, Desnos
et al. 1996 ). Furthermore, multitemporal SAR or SAR / VIR combinations are used
together with topographic maps ( Bonansea 1995, Brakenridge 1995, Kannen et al.
1995, Otten and Persie 1995 ).

5.4. Ice /snow monitoring

The fusion of data in the ® eld of ice monitoring provides results with higher
reliability and more detail ( Ramseier et al. 1993, Armour et al. 1994 ). Regarding the
use of SAR from di€ erent orbits for snow monitoring the amount of distorted areas
due to layover, shadow and foreshortening can be reduced signi® cantly ( Haefner
et al. 1993 ).

5.5. Geology
Multisensor image fusion is well implemented in the ® eld of geology and a widely
applied technique for geological mapping. It is a well known fact that the use of
multisensor data improves the interpretation capabilities of the images. Geological
 Multisensor image f usion in remote sensing 841

features which are not visible in the single data alone are detected from integrated
imagery. In most cases VIR is combined with SAR based on the fact that the data
sets complement each other. They introduce information on soil geochemistry,
vegetation and land use ( VIR) as well as soil moisture, topography and surface
roughness (SAR) ( Daily et al . 1979, Blom and Daily 1982, Rebillard and Nguyen
1982, Reimchen 1982, Haydn et al . 1982, Aarnisalo 1984, Evans 1988, Baker and
Henderson 1988, Hopkins et al . 1988, Paradella et al . 1988 and Taranik 1988, Guo
and Pinliang 1989, Harris and Murray 1989, Koopmans et al . 1989, Harris et al .
1990, Grasso 1993, Jutz and Chorowicz 1993, Koopmans and Forero 1993,
Koopmans and Richetti 1993, Ye sou et al . 1993 a, 1993 b, 1994, Ray et al . 1995).

6. Advantages and limitations

In order to summarise the review some issues related to advantages and
disadvantages along with constraints of image fusion are discussed below.

6.1. Selection of appropriate data set and techniques

The decision on which technique is the most suitable is very much driven by
the application. Therefore, it is very di cult to make general statements on the
quality of a fusion technique. In addition the type of data available and statistical
measurements, e.g., correlation matrix may support the choice.
The choice of a data set is application dependent as is the technique used to fuse
the data ( Pohl 1996). Other limiting factors in respect to the data selection are
operational constraints of the sensors that are supposed to acquire the data, including
receiving stations and data distributors ( Pohl 1996, Polidori and Mangolini 1996).
The characteristics of fused image data depend very much on the applied pre-
processing and the image fusion technique. The constraints are related to the disturb-
ance of spectral content of the input data and a blurring e€ ect when introducing
images with a low SNR. Also, the temporal aspect should not be underestimated.
Changes in the area between the acquisition dates of the imagery might introduce
problems in the fused image. The user of multisensor data in the context of image
fusion has to be aware of the physical characteristics of the input data in order to
be able to select appropriate processing methods and to judge the resulting data
( Pohl 1996 ).

6.1.1. Pre-processing
All sensor-speci® c corrections and enhancements of image data have to be applied
prior to image fusion since the techniques refer to sensor-speci® c e€ ects. After image
fusion the contribution of each sensor cannot be distinguished or quanti® ed in order
to be treated accordingly. A general rule of thumb is to ® rst produce the best single-
sensor geometry and radiometry (geocoding, ® lter, line and edge detection, etc.) and
then fuse the images. Any spatial enhancement performed prior to image fusion will
bene® t the resulting fused image. An advantage is the possibility of ® ltering and
enhancing the data during the geocoding process to avoid multiple resampling. The
data has to be resampled at the pixel spacing required for the desired image fusion
( Pohl 1996 ).
The importance of geometric accuracy to avoid artefacts and misinterpretation
in pixel based image fusion should not be underestimated. Pixels registered to each
other should refer to the same object on the ground. This implies that the data
should be geocoded with sub-pixel accuracy. The DEM therefore plays an important
842 C. Pohl and J. L . van Genderen

role in this process. The need for DEMs of high quality and appropriate grid spacing
is evident.

6.1.2. T echniques and image combinations

Categorized by technique the following sections summarize observations relevant
to the judgement of the usefulness of a certain fusion technique:

6.1.2.1. RGB
Digital numbers in single images in¯ uence the colours in the RGB composite.
This implies that the following factors need consideration:

1. Histogram stretching of single channels in¯ uences the visibility of features in

the ® nal colour composite.
2. Inversion of image channels might be desirable to assign colour to features.
3. RGB channel assignment signi® cantly in¯ uences the visibility of features and
visual interpretation by a human interpreter ( blue = water, green = land, etc.).

The technique is simple and does not require CPU time-intensive computations.
RGB overlay protects the contribution from optical imagery from being greatly
a€ ected by speckle from SAR ( Pohl 1996).

6.1.2.2. Band combinations

The use of linear combinations have the disadvantage that these band
combinations lead to a high correlation of the resulting bands and that part of the
spectral information is lost compared to the original multispectral image ( Ehlers
1987, 1991). The fusion of SAR with VIR data improves the interpretation of the
SAR data. Subsequently, it helps applications that bene® t from the interpretation of
up-to-date SAR data (urban growth, coastal zone monitoring, tidal activities, soil
moisture studies, etc.). The resulting fused image depends very much on the appear-
ance and content of the SAR data. As a result, the SAR data have to be selected
according to the ® eld of interest of the application. In¯ uence of terrain on SAR
backscatter reduces the suitability of the data for recognizing features such as roads,
cities, rivers, etc. This type of technique does not solve the cloud-cover problem
because the range of digital numbers corresponding to clouds is preserved and even
enhanced if not excluded from the calculation ( Pohl 1996).

6.1.2.3. Brovey
The spectral content of the VIR data is preserved while introducing the texture
from SAR. The resulting image is not quite as sharp as the one produced from
multiplication only. The water-land boundaries are well de® ned in the fused images;
it allows colours to be assigned to the water currents (e.g., tidal inlets) ( Pohl 1996).

6.1.2.4. PCA
Radiometric pre-processing plays an important role in relation to the spectral
content of the fused image. The appearance of SAR signi® cantly in¯ uences the feature
visibility in the fused VIR/ SAR image. As a consequence, features that are detectable
on SAR data can be introduced to the VIR data by image fusion to complement
the data (e.g., soil moisture, urban area, oceanographic objects) ( Pohl 1996). The
PCA technique enables the integration of more than three types of data ( Zobrist
et al . 1979 ).
 Multisensor image f usion in remote sensing 843

Principal component SAR images show potential for topographic mapping. This
is valid, in particular, for the three-dimensional impression of topography and change
detection. The possibilities have not yet been fully explored, e.g., combination of
principal components with optical data ( Pohl 1996 ).

6.1.2.5. IHS
The IHS o € ers a controlled visual presentation of the data. The informative
aspects are presented in IHS using readily identi® able and quanti® able colour
attributes that can be distinctly perceived. Numerical variations can be uniformly
represented in an easily perceived range of colours ( Harris et al . 1990 ). Other
techniques produce images which are di cult to interpret quantitatively and
qualitatively because the statistical properties have been manipulated and the original
integrity of the data is disturbed. Related to the IHS technique the hue has to be
carefully controlled since it associates meaningful colour with well de® ned charac-
teristics of the input. An advantage of the IHS versus the RGB is the ability of
integrating four instead of three channels in the fused image ( Rothery and Francis
1987 ). IHS fused imagery have the capability of allocating data from the SAR to
cloud covered areas without having to identify the clouds at an earlier stage. The
speckle is preserved from the SAR data in the fused image. It shows similarities with
the Brovey transformation in terms of spectral content of the imagery. A disadvantage
is the reduced spatial detail compared to original optical data ( Pohl 1996).

6.1.2.6. Mosaic
The mosaic has an important position amongst the image fusion techniques as
far as cloud removal from VIR and the replacement of radiometrically distorted
SAR data is concerned. The result depends very much on the quality of the mask
designed for the mosaic. This is a critical point for the optical imagery. The
identi® cation of foreshortening, layover and shadow areas in the SAR is based on
DEM calculations and pure geometry. These products are often delivered with the
SAR image itself. It is essential to match the histograms of the various input data
to each other. It can be used in combination with any other image fusion technique
( Pohl 1996 ).

6.2. Assessment criteria

Working in the ® eld of multisensor image fusion, the evaluation of the achieved
results becomes relatively complex because of the di€ erent sources of data that are
involved. The di€ erent aspects of image acquisition of the various sensors have to
be considered as well as the approach of the image fusion itself plays a role.
A numerical quality assessment of image fusion processes was recently introduced
by Ranchin et al. ( 1996 ). They implemented mathematical conditions to judge the
quality of merged imagery in respect to their improvement of spatial resolution while
preserving the spectral content of the data. With the use of di€ erence images a
comparison was made with the original data introduced to the fusion process. In
addition, image bias, variance and correlation of imagery were taken into account
as indicators during the mathematical quality assessment ( Ranchin et al. 1996 ).
Another possibility is to validate ® ndings from fused data by testing the methods
on simulated data ( known parameters) followed by a comparison with actual data
sets. It is necessary to have access to a variety of situations, relevant parameters and
ground truth ( Polidori and Mangolini 1996 ).
844 C. Pohl and J. L . van Genderen

In this respect much more research is required to provide objective evaluation

methods for fused imagery ( Pohl 1996).

6.3. L evels of image f usion

Some considerations regarding the judgement of image fusion are necessary
regarding the processing levels as mentioned above. A constraint of pixel based
fusion is the necessity of accurate geocoding, including a resampling of the data. The
geometric correction requires knowledge on the sensor viewing parameters along
with software that takes into account the image acquisition geometry. In areas with
considerable height variations a digital elevation model ( DEM ) is necessary. This is
especially important for SAR data processing due to the side-looking geometry of
the sensor. In addition, pixel based methods deal with very large data volumes and
long computation times are involved. Frequently, the di culty of comparing pixels
from heterogeneous data is mentioned in the literature. Feature level image fusion
uses corresponding structures which makes the geometry a less critical issue. The
pixel based methods have the advantage against feature or information level fusion
of using the most original data. It avoids a loss of information which occurs during
a feature extraction process (Mangolini 1994 ).

6.4. Operational image f usion

The fusion of multisensor data is limited by several factors. Often, the lack of
simultaneously acquired multisensor data hinders the successful implementation of
image fusion. In the case of large di€ erences in spatial resolution of the input data
problems arise from the limited (spatial) compatibility of the data. Since there is no
standard procedure of selecting the optimal data set, the user is often forced to work
empirically to ® nd the best result. The use of multisensor data requires more skills of
the user in order to handle the data appropriately which also invokes the availability
of software that is able to process the data. The validation of the fusion algorithm
bares some insecurity ( Polidori and Mangolini 1996 ). Furthermore, the lack of
available ground truth ( Lozano-Garcia and Ho€ er 1993) forms another constraint,
and of course of the use of more than one scene has an impact on the costs involved.

6.5. Future improvements

The development of powerful hardware along with more and more sophisticated
software allow the implementation of algorithms and techniques that require large
data volumes and time intensive computation such as the wavelet approach. An
improvement and the development of new techniques is currently being developed,
e.g., concerning ANN and the wavelet transform as already mentioned in the context
of the ARSIS method ( Ranchin and Wald 1993). A combination of wavelet transform
and ANN classi® cation was already presented by Melis and Lazzari ( 1994). Also,
the involvement of expert systems in a GIS can support the integration and evaluation
of fused image data.
A very important aspect in future research and development activities is to place
emphasise on methods to estimate and assess the quality of fused imagery. A start
is the mathematical approach along with a set of criteria to check the radiometric
integrity of the fused imagery described by Mangolini et al. ( 1993) and Ranchin
et al. ( 1996 ).
 Multisensor image f usion in remote sensing 845

More ¯ exible satellite sensors are expected to contribute to the optimisation of

the input data set (e.g., Radarsat). The operation of ERS-1 and ERS-2 in the so-called
`Tandem-mode’ provides a solution to problems that occur due to temporal changes
of ground cover between the acquisition dates of SAR data. Future satellites with
multiple sensors on board will provide the opportunity to simultaneously acquire a
data set. An example is SPOT-4 with the VEGETATION program ( 1 km resolution)
as well as multispectral HRV at 20 m resolution ( Pohl 1996). Value-added products
are to be provided by the data processing agencies along with special training of the
data users ( Polidori and Mangolini 1996 ).
With upcoming improved sensors, the importance of multisensor image fusion
and interpretation will increase and open new possibilities in Earth observation
for operational applications for the protection and development of the Earth’s
environment ( Pohl 1996).

Acknowledgments
The authors thank Dr H. Ye sou for his valuable comments and suggestions
on the ® rst draft of the manuscript. ITC’s research on image fusion has been
externally funded by the CEC (CHRX-CT93-0310 ), EOSAT, ESA, EURIMAGE,
EUROSENSE, MCS, MPA, NLR and NASDA. Their support to this research is
gratefully acknowledged by the authors.

Appendix AÐ Abbreviations and acronyms

ANN Arti® cial Neural Network
ARSIS ame lioration de la re solution spatial par injection de structures
CAMS Calibrated Airborne Multispectral Scanner (CAMS)
COS COmponent Substitution
CRT Cathode Ray Tube
DN Digital Number
HPF High Pass Filtering
HSV Hue Saturation Value colour representation
IHS Intensity Hue Saturation colour space
LUT Look Up Table
MRA MultiResolution Approach
OIF Optimum Index Factor
PAN PANchromatic band
PC1 First Principal Component
PCA Principal Component Analysis
PCS Principal Component Substitution
RGB Red Green Blue colour space
RVS Regression Variable Substitution
SAR Synthetic Aperture Radar
SIR-B Shuttle Imaging Radar Experiment B
SNR Signal to Noise Ratio
SPOT System d’Observation de la Terre
SPS Standardised PC Substitution
TIMS Thermal Infrared Multispectral Scanner ( TIMS)
TM Landsat Thematic Mapper
VIR Visible and Infrared
XS SPOT multispectral bands ( 3)
YIQ Luminance Chrominance colour space
846 C. Pohl and J. L . van Genderen

References
A arnisalo, J ., 1984 , Image processing and integration of geophysical, MSS and other data
as a tool for numerical exploration in glaciated Precambrian terrain. Proceedings of
the International Symposium on Remote Sensing of Environment, Proceedings of T hird
T hematic Conference, Remote Sensing for Exploration Geology, Colorado Springs
(Colorado Springs: ISRS), pp. 107± 128.
A hern, F . J ., G oodenough, D . G ., G rey, A . L ., and R yerson, R . A ., 1978, Simultaneous
microwave and optical wavelength observations of agricultural targets. Canadian
Journal of Remote Sensing, 4, 127± 142.
A lbertz, J ., and T auch, R ., 1994, Mapping from space Ð Cartographic applications of satellite
image data. GeoJournal , 32, 29± 37.
A rmour, B ., E hrismann, J ., C hen, F ., and B owman, G ., 1994, An integrated package for
the processing and analysis of SAR imagery, and the fusion of radar and passive
microwave data. In Proceedings ISPRS Symposium, 6± 10 June 1994, Ottawa, Canada
(Ottawa: International Archives of Photogrammetry and Remote Sensing ),
pp. 299± 306.
A schbacher, J ., and L ichtenegger, J ., 1990, Complementary nature of SAR and optical
data: a case study in the Tropics. Earth Observation Quarterly, 31, 4± 8.
A schbacher, J ., G iri, C . P ., O fren, R . S ., T iangco, P . N ., D elso, J . P ., S uselo, T . B .,
V ibulsresth, S ., and C harupat, T ., 1994, Tropical mangrove vegetation mapping
using advanced remote sensing and GIS technology. Executive Summary, ADC Austria
and Asian Institute of T echnology (Bangkok: AIT).
B adji, M ., 1995, Flood plain management in Wallonia. In Application Achievements of ERS-1,
New V iews of the Earth, Preliminary V ersion (Paris: European Space Agency),
pp. 114± 115.
B aker, R . N ., and H enderson, F . B ., 1988, Results of the GEOSAT programme to
evaluate SPOT data for petroleum and mineral exploration. SPOT 1: Image utilization,
assessment, results, Proceedings CNES Conference, Paris, France, Cepadue s Editions
( Paris: CNES), pp. 731± 738.
B lom, R . G ., and D aily, M ., 1982, Radar image processing for rock type discrimination.
I.E.E.E. T ransactions on Geoscience and Remote Sensing, 20, 343± 351.
B loom, A ., F ielding, E ., and F u, X ., 1988, A demonstration of stereophotogrammetry with
combined SIR-B and Landsat-TM images. International Journal of Remote Sensing, 9,
1023± 1038.
B onansea, E ., 1995, Flood monitoring and assessment in Piemonte. In Application
Achievements of ERS-1, New V iews of the Earth ( Paris: European Space Agency),
SP-1176 / 11, p. 95.
B rakenridge, G . R ., 1995, Flood mapping in the Mississippi. In Application Achievements of
ERS-1, New V iews of the Earth ( Paris: European Space Agency), SP-1176 / 11, p. 92.
B risco, B ., and B rown, R . J ., 1995, Multidate SAR / TM synergism for crop classi® cation in
Western Canada. Photogrammetric Engineering and Remote Sensing, 61, 1009± 1014.
B risco, B ., U laby, F . T ., and D obson, M . C ., 1983, Spaceborne SAR data for land cover
classi® cation and change detection. Digest of International Geoscience and Remote
Sensing Symposium , 1, 1.1± 1.8.
B uchroithner, M . F ., R aggam, J ., S trobl, D ., and A lmer, A ., 1992, Multisensor data
merging for digital landscape models. Analysis of Earth Observation Space Data
Integration, Proceedings EARSeL W orkshop, held in Sinaia, Romania, 28± 30 October
1992 (Paris: EARSeL ), pp. 34± 42.
C ampbell, N . A ., 1993, Towards more quantitative extraction of information from remotely
sensed data. Advanced Remote Sensing, Conference Proceedings, held in Sydney,
Australia, 2, 29± 40.
C arper, W . J ., L illesand, T . M ., and K ieffer, R . W ., 1990, The use of Intensity-Hue-
Saturation transformations for merging SPOT Panchromatic and multispectral image
data. Photogrammetric Engineering and Remote Sensing, 56, 459± 467.
C havez, P . S ., 1987, Digital merging of Landsat TM and digitized NHAP data for 1 : 24 000
scale image mapping. Photogrammetric Engineering and Remote Sensing, 52, 1637± 1646.
C havez, P . S ., B erlin, G . L ., and S owers, L . B ., 1982, Statistical methods for selecting
Landsat MSS ratios, Journal of Applied Photographic Engineering , 8, 23± 30.
 Multisensor image f usion in remote sensing 847

C havez, P . S ., G uptill, S . C ., and B owell, J . H ., 1984, Image Processing Techniques for

TM data. T echnical Papers, Proceedings 50th Annual Meeting of ASPRS, Washington
DC, U.S.A. , 2, 728± 742.
C havez, P . S ., S ides, S . C ., and A nderson, J . A ., 1991, Comparison of three di € erent methods
to merge multiresolution and multispectral data: TM & SPOT pan. Photogrammetric
Engineering and Remote Sensing, 57, 295± 303.
C heng, P ., T outin, T ., and P ohl, C ., 1995, A comparison of geometric models for multi-
source data fusion. Remote Sensing, GIS and GPS in Sustainable Development and
Environmental Monitoring, GeoInformatics ’95, Proceedings of International Symposium,
26± 28 May 1995, Hong Kong (Hong Kong: Chinese University of Hong Kong ),
pp. 11± 17.
C hiesa, C . C ., and T yler, W . A ., 1990, Data fusion of o € -nadir SPOT panchromatic images
with other digital data sources. T echnical Papers 1990, ACSM-ASPRS Annual
Convention, Image Processing and Remote Sensing, 4, 86± 98.
C hiuderi, A ., and F ini, S ., 1993, Data fusion techniques for remotely sensed images. T ime-
varying Image Processing and Moving Object Recognition, Proceedings of 4th W orkshop,
Florence, Italy, 10± 11 June 1993 , pp. 1± 8.
C liche, G ., B onn, F ., and T eillet, P ., 1985, Integration of the SPOT Pan channel into its
multispectral mode for image sharpness enhancement. Photogrammetric Engineering
and Remote Sensing, 51, 311± 316.
C omhaire, A . L ., O zer, A ., and J aspar, A ., 1994, The contribution of ERS-1 and SPOT
data in the morphologic and bathymetric study of the north sea and the Scheldt
Estuary. Proceedings of First ERS-1 Pilot Project W orkshop, T oledo, Spain, 22± 24
June 1994 , SP-365 (Paris: European Space Agency), pp. 297± 302.
C orves, C ., 1994, Assessment of multi-temporal ERS-1 SAR Landsat TM data for mapping
the Amazon river ¯ oodplain. Proceedings of First ERS-1 Pilot Project W orkshop,
T oledo, Spain, 22± 24 June 1994 , SP-365 (Paris: European Space Agency), pp. 129± 132.
C rist, E . P ., 1984, Comparison of coincident Landsat-4 MSS and TM data over an agricultural
region. T echnical Papers, Proceedings 50th Annual Meeting ASP-ACSM Symposium,
ASPRS, Washington DC, U.S.A. (Washington: ASPRS), pp. 508± 517.
D aily, M ., 1983, Hue-saturation-intensity split-spectrum processing of SEASAT radar imagery.
Photogrammetric Engineering and Remote Sensing, 49, 349± 355.
D aily, M . I ., F arr, T ., and E lachi, C ., 1979, Geologic interpretation from composited
radar and Landsat imagery. Photogrammetric Engineering and Remote Sensing, 45,
1009± 1116.
D allemand, J . F ., L ichtenegger, J ., K aufmann, V ., P audyal, D . R ., and R eichert, A .,
1992, Combined analysis of ERS-1 SAR and visible /infrared RS data for land
cover/ land use mapping in tropical zone: a case study in Guinea. Space at the Service
of our Environment, Proceedings First ERS-1 Symposium, Cannes, France, 6± 8
November 1992, SP-359 (Paris: European Space Agency), pp. 555± 561.
D e G roof, H ., S owter, A ., and S ieber, A . J ., 1992, The TREES project: Creating an
environment for the routine classi® cation and image processing of ERS-1 SAR imagery.
International Space Year: Space Remote Sensing, Proceedings 12th Annual I.E.E.E.
International Geoscience and Remote Sensing Symposium (IGARSS ’92), Houston,
U.S.A. , edited by R. Williamson with collaboration of T. Stein (New York: I.E.E.E.),
pp. 1343± 1345.
D esnos, Y .-L ., M ayer, T ., and S ardar, A . M ., 1996, Multitemporal ERS-1 SAR images of
the Brahmaputra ¯ ood plains in Northern Bangladesh. Earth Observation Quarterly,
51, 6± 10.
D jamdji, J .-P ., B ijaoui, A ., and M anieri, R ., 1993, Geometrical registration of images: The
multiresolution approach. Photogrammetric Engineering and Remote Sensing, 59,
645± 653.
D omik, G ., L eberl, F ., and C imino, J . B ., 1986, Multi incidence angle SIR-B experiment
over Argentina: Generation of secondary image products. I.E.E.E. T ransactions on
Geoscience and Remote Sensing, 24, 492± 497.
D uguay, G ., H older, G ., H owarth, P ., and L eD rew, E ., 1987, Integrating remotely sensed
data from di € erent sensors for change detection. Proceedings of the I.E.E.E.
International Geoscience and Remote Sensing Symposium (IGARSS ’87), Ann Arbor,
U.S.A., 18± 21 May 1987 ( New York: I.E.E.E.), p. 333± 000.
848 C. Pohl and J. L . van Genderen

E hlers, M ., 1987 , Integrative Auswertung von digitalen Bildern aus der Satelliten-
photogrammetrie und-fernerkundung im Rahmen von geographischen Informations-
systemen. Wissenschaftliche Arbeiten der Fachrichtung Vermessungswesen der
UniversitaÈ t Hanover, Germany, Dissertation, No. 149.
E hlers, M ., 1991, Multisensor image fusion techniques in remote sensing, ISPRS Journal of
Photogrammetry and Remote Sensing, 46, 19± 30.
E hlers, M ., 1993, Integration of GIS, remote sensing, photogrammetry and cartography: the
geoinformatics approach. GIS GEO-Informations-Systeme , 6, 18± 23.
ESA, 1995 a, ERS-1 SAR images of ¯ ood in Northern Europe (January 1995). Earth
Observation Quarterly, 47, 13.
ESA, 1995 b, Multitemporal image Maastricht ( The Netherlands). Earth Observation Quarterly,
47, 14.
E ssadiki, M ., 1987, A combination of panchromatic and multispectral SPOT images for
topographic mapping. IT C Journal , 1987-1, 59± 66.
E vans, D ., 1988, Multisensor classi® cation of sedimentary rocks. Remote Sensing of
Environment , 25, 129± 144.
F ellah, K ., and T holey, N ., 1996, Flood risk evaluation and monitoring in North East
France. In Application Achievements of ERS-1, New V iews of the Earth, Preliminary
V ersion ( Paris: European Space Agency), pp. 116± 119.
F ranklin, S . E ., and B lodgett, C . F ., 1993, An example of satellite multisensor data fusion.
Computers and Geoscience , 19, 577± 583.
F ung, T ., and L eD rew, E ., 1987, Application of Principal Component Analysis to change
detection. Photogrammetric Engineering and Remote Sensing, 53, 1649± 58.
G enderen, J . L . van, and P ohl, C ., 1994, Image fusion: Issues, techniques and applications.
Intelligent Image Fusion, Proceedings EARSeL W orkshop, Strasbourg, France, 11
September 1994 , edited by J. L. van Genderen and V. Cappellini ( Enschede: ITC),
pp. 18± 26.
G enderen, J . L . van, C racknell, A . P ., K onecny, G ., and S ieber, A ., 1994, Synergy of
remotely sensed data: European scienti® c network in the ® eld of remote sensing.
Proceedings of the First ERS-1 Pilot Project W orkshop, T oledo, Spain, 22± 24 June
1994 , SP-365 ( Paris: European Space Agency), pp. 229± 234.
G illespie, A . R ., K ahle, A . B ., and W alker, R . E ., 1986, Colour enhancement of highly
correlated images. I. Decorrelation and HSI contrast stretches. Remote Sensing and
Environment , 20, 209± 235.
G rasso, D . N ., 1993, Application of the IHS colour transformation for 1 : 24 000-scale geologic
mapping: a low cost SPOT alternative. Photogrammetric Engineering and Remote
Sensing, 59, 73± 80.
G riffiths, G . H ., 1988, Monitoring urban change from Landsat TM and SPOT satellite
imagery by image di € erencing. Proceedings I.E.E.E. International Geoscience and
Remote Sensing Symposium (IGARSS ’88), Edinburgh, Scotland, 13± 16 September 1988
( New York: I.E.E.E.), pp. 493± 497.
G uo, H . D ., and P inliang, D ., 1989, Integrated MSS-SAR-SPOT-geophysical geochemical
data for exploration geology in Yeder area. CAS-IRSA, pp. 1± 8.
G uyenne, T .-D ., 1995, ERS-1 /2 multitemporal tandem image: Helsinki (Finland ). Supplement
to Earth Observation Quarterly, 48, 4.
H aack, B ., and S lonecker, T ., 1991, Sensor fusion for locating villages in Sudan. GIS /L IS
ACSM-ASPRS Fall Convention, T echnical Papers, Atlanta, U.S.A., 28 October 1991
(Atlanta: ASPRS), pp. B97± B107.
H aefner, H ., H olecz, F ., M eier, E ., N üesch D ., and P iesbergen, J ., 1993, Capabilities and
limitations of ERS-1 SAR data for snowcover determination in mountainous regions.
Space at the Service of our Environment, Proceedings of the Second ERS-1 Symposium,
Hamburg, Germany, 11± 14 October 1993 , SP-361 ( Paris: European Space Agency),
pp. 971± 976.
H all, D . L ., 1992, Mathematical techniques in multisensor data f usion (Norwood: Artech
House Inc.).
H arris, G . H ., and G raham, L . C ., 1976, Landsat-radar synergism. Proceedings of the XII
Congress of ISP, Helsinki, Finland.
H arris, J . R ., and M urray, R ., 1989, IHS transform for the integration of radar imagery
 Multisensor image f usion in remote sensing 849

with geophysical data. Proceedings of the 12th Canadian Symposium on Remote Sensing,
V ancouver, Canada ( Vancouver: CSRS), pp. 923± 926.
H arris, J . R ., M urray, R ., and H irose, T ., 1990, IHS transform for the integration of radar
imagery with other remotely sensed data. Photogrammetric Engineering and Remote
Sensing, 56, 1631± 1641.
H arrison, B . A ., and J upp, D . L . B ., 1990, Introduction to image processing. MicroBRIAN
Resource Manual, Part 2 (Melbourne: CSRIO Publications).
H aydn, R ., D alke, G . W ., H enkel, J ., and B are, J . C ., 1982, Application of the IHS color
transform to the processing of multisensor data and image enhancement. Remote
Sensing of Arid and Semi-Arid L ands, Proceedings of International Symposium (Cairo:
ISRS), pp. 599± 607.
H inse, M ., and C oulombe, A ., 1994, Le radar ¼ seul ou avec d’autres?. Numerimage, Bulletin
d ’Information Quadrimestriel (Quebec) , 3, 2± 4.
H inse, M ., and P roulx, J ., 1995, La spatiocarte regionale. Numerimage, Bulletin d’Information
Quadrimestriel (Quebec) , 3, p. 2.
H opkins, H . R ., N avail, H ., B erger, Z ., M erembeck, B . F ., B rovey, R . L ., and
S chriver, J . S ., 1988, Structural analysis of the Jura mountains± Rhine Graben inter-
section for petroleum exploration using SPOT stereoscopic data. SPOT 1: Image
utilization, assessment, results, Proceedings CNES Conference, Paris, France, Cepadues
Editions (Paris: CNES), pp. 803± 810.
H ussin, Y . A ., and S haker, S . R ., 1996, Optical and radar satellite image fusion techniques
and their applications in monitoring natural resources and land use change. Synthetic
Aperture Radar, Proceedings European Conference (EUSAR ’96), KoÈnigswinter,
Germany, 26± 28 March 1996 (Berlin: VDE), pp. 451± 456.
J anssen, L . L . F ., J aarsma, M . N ., and L inden, E . T . M . van der , 1990, Integrating
topographic data with remote sensing for land-cover classi® cation. Photogrammetric
Engineering and Remote Sensing, 56, 1503± 1506.
J utz, S . L ., and C horowicz, J ., 1993, Geological mapping and detection of oblique extension
structures in the Kenyan Rift Valley with a SPOT/ Landsat-TM datamerge.
International Journal of Remote Sensing, 14, 1677± 1688.
K achhwalha, T . S ., 1993, Temporal and multisensor approach in forest /vegetation mapping
and corridor identi® cation for e € ective management of Rajaji Nat. Park, Uttar
Pradesh, India. International Journal of Remote Sensing, 14, 3105± 3114.
K annen, A ., O berstadler, R ., and P rotman, F ., 1995, Flood mapping in Germany.
Application Achievements of ERS-1, New V iews of the Earth, Preliminary V ersion (Paris:
European Space Agency), pp. 124± 125.
K attenborn, G ., N ezry, E ., D e G randi, G ., and S ieber, A . J ., 1993, High resolution detection
and monitoring of changes using ERS-1 time series. Space at the Service of our
Environment, Proceedings of the Second ERS-1 Symposium, 11± 14 October, Hamburg,
Germany, I ( Paris: European Space Agency), pp. 635± 642.
K aufmann, K .-H ., and B uchroithner, M . F ., 1994, Herstellung und Anwendungs-
moÈ glichkeiten von Satellitenkarten durch digitale Kombination von Landsat-TM und
KWR-1000-Daten. Zeitschrift f uÈ r Photogrammetrie und Fernerkundung , 62, 133± 137.
K eys, L . D ., S chmidt, N . J ., and P hillips, B . E ., 1990, A prototype example of sensor fusion
used for a siting analysis. T echnical Papers 1990, ACSM-ASPRS Annual Convention,
Image Processing and Remote Sensing, 4, 238± 249.
K nipp, K .-H ., 1993, Speckle reduction applied to ERS-1 SAR images. International Mapping
f rom Space, Proceedings ISPRS Working Group IV / 2 W orkshop and Conference,
Hanover, Germany, 27 September± 1 October 1993 , 15, 177± 190.
K ohl, H .-G ., N ezry, E ., and D e G roof, H ., 1994, Crop acreage estimation with ERS-1 PRI
images. Earth Observation Quarterly, 46, 6± 9.
K oopmans, B . N ., and F orero, G . R ., 1993, Airborne SAR and Landsat MSS as com-
plementary information source for geological hazard mapping. ISPRS Journal of
Photogrammetry and Remote Sensing, 48, 28± 37.
K oopmans, B . N ., and R ichetti, E ., 1993, Optimal geological data extraction from SPOT-
Radar synergism with samples from Djebel Amour (Algeria), Red Sea Hills (Sudan),
Sirte Basin ( Libya) and Magdalena Valley (Colombia). From Optics to Radar, SPOT
and ERS Applications, Conference Proceedings, 10± 13 May 1993, Paris, France (Paris:
Cepad ), pp. 263± 274.
850 C. Pohl and J. L . van Genderen

K oopmans, B . N ., B akx, J . P . G ., D ijk, P . M . van, and A l F asatwi, Y . A ., 1989, SPOT-

RADAR synergism for geological mapping in arid areas. Project Report Prosper
Project, Final Report, CNES-SPOT Image-ITC.
L eckie, D . G ., 1990 , Synergism of SAR and visible /infrared data for forest type discrimination.
Photogrammetric Engineering and Remote Sensing, 56, 1237± 1246.
L ichtenegger, J ., 1991, Combining optical /infrared and SAR images for improved remote
sensing interpretation. ESA Bulletin, 66, 119± 121.
L ichtenegger, J ., and C alabresi, G ., 1995, ERS monitors ¯ ooding in The Netherlands. ESA
Bulletin, 81, Title page & p. 94.
L ichtenegger, J ., D allemand, J .-F ., R eichert, P ., R ebillard, P ., and B uchroithner, M . ,
1991, Multi-sensor analysis for land use mapping in Tunisia. Earth Observation
Quarterly , 33, 1± 6.
L oercher, G ., and W ever, T ., 1994, Integrated use of optical and radar data from the MAC-
Europe campaign 1991 for geological mapping of the Hekla area (Iceland ). Proceedings
First International Airborne Remote Sensing Conference and Exhibition, Strasbourg,
France, 11± 15 September 1994 (Strasbourg: ERIM), pp. III-471± III-478.
L ozano-G arcia, D . F ., and H offer, R . M ., 1993, Synergistic e € ects of combined Landsat-TM
and SIR-B data for forest resources assessment. International Journal of Remote Sensing,
14, 2677± 2694.
M ac Intosh, H ., and P rofeti, G ., 1995, The use of ERS SAR data to manage ¯ ood emergencies
at a smaller scale. Proceedings of Second ERS Applications Workshop, L ondon, U.K.,
6 ± 8 December 1995 (Paris: ESA publication), pp. 243± 246.
M allat, S . G ., 1989, A theory for multiresolution signal decomposition: The wavelet
representation. I.E.E.E. T ransactions on Pattern Analysis and Machine Intelligence,
11, 674± 693.
M angolini, M ., 1994, Apport de la fusion d’images satellitaires multicapteurs au niveau pixel
en te le de tection et photo-interpre tation. Dissertation published at the University of
Nice± Sophia Antipolis, France, 15 November 1994.
M angolini, M ., and A rino, O ., 1996 a, Improvement of agriculture monitoring by the use
of ERS /SAR and Landsat / TM. Fusion of Earth Data, Proceedings EARSeL Conference,
Cannes, France, 6± 8 February 1996 ( Paris: EARSeL ), pp. 29± 35.
M angolini, M ., and A rino, O ., 1996 b, ERS-1SAR and Landsat-TM multitemporal fusion
for crop statistics. Earth Observation Quarterly, 51, 11± 15.
M angolini, M ., R anchin, T ., and W ald, L ., 1993, Fusion d’images SPOT multispectrale
( XS) et panchromatique (P), et d’images radar. From Optics to Radar, SPOT and ERS
Applications, Conference Proceedings, 10± 13 May 1993, Paris, France ( Paris: Cepad ),
pp. 199± 209.
M arek, K .-H ., and S chmidt, K ., 1994, Preliminary results of the comparative analysis of
ERS-1 and ALMAZ-1 SAR data. ISPRS Journal of Photogrammetry and Remote
Sensing, 49, 12± 18.
M atte, R ., 1994, L’ortho-image au service de la cartographie. Numerimage, Bulletin
d ’Information Quadrimetriel (Quebec) , 3, 9± 12.
M atthews, J ., and G affney, S ., 1994, A study of ¯ ooding on Lough Corrib, Ireland during
early 1994 using ERS-1 SAR data. Proceedings of First ERS-1 Pilot Project Workshop,
T oledo, Spain, 22± 24 June 1994 , SP-365 (Paris: European Space Agency), pp. 125± 128.
M elis, M ., and L azzari, A ., 1994, Multitemporal and single image features extraction, from
ERS-1 SAR image data, with wavelet transform unsupervised neural networks.
Proceedings of First ERS-1 Pilot Project W orkshop, T oledo, Spain, 22± 24 June 1994 ,
SP-365 ( Paris: European Space Agency), pp. 493± 500.
M itiche, A ., and A ggarwal, J . K ., 1986, Multiple sensor integration / fusion through image
processing: a review. Optical Engineering , 25, 380± 386.
M oran, M . S ., 1990. A Window-based technique for combining Landsat Thematic Mapper
thermal data with higher-resolution multispectral data over agricultural land.
Photogrammetric Engineering and Remote Sensing, 56, 337± 342.
M ouat, D . A ., M ahin, G . G ., and L ancaster, J ., 1993, Remote sensing techniques in the
analysis of change detection. Geocarto International , 2, 39± 50.
M unechika, C . K ., W arnick, J . S ., S alvaggio, C ., and S chott, J . R ., 1993, Resolution
enhancement of multispectral image data to improve classi® cation accuracy.
Photogrammetric Engineering and Remote Sensing, 59, 67± 72.
 Multisensor image f usion in remote sensing 851

N andhakumar, N ., 1990, A phenomenologica l approach to multisource data integration:

analyzing infrared and visible data. Multisource Data Integration in Remote Sensing,
Proceedings of W orkshop, Maryland, U.S.A., 14± 15 June 1990 , NASA Conference
Publication 3099 (Greenbelt: NASA), pp. 61± 73.
N ezry, E ., M ougin, E ., L opes, A ., and G astellu- E tchegorry, J . P ., 1993, Tropical vegetation
mapping with combined visible and SAR spaceborne data. International Journal of
Remote Sensing, 14, 2165± 2184.
O prescu, N ., S erban, F ., B raescu, C ., and N ecsoiu, M ., 1994, Technological researches
and developments, regarding the integration of SAR-ERS data with other types of
satellite and ground information, as well as determining their complementarity, with
applications over the continental platform of the Black Sea, the Danube Delta and
the seismic zone Vrancea; some contributions to data calibration and validation.
Proceedings of First ERS-1 Pilot Project W orkshop, T oledo, Spain, 22± 24 June 1994 ,
SP-365 ( Paris: European Space Agency), pp. 199± 204.
O tten, M . P . G ., and P ersie, M . van, 1995, SAR image of ¯ oods using PHARS system.
Earth Observation Quarterly, 47, 15.
P aradella, W . R ., V itorello, I ., L iu, C . C ., M attos, J . T ., M eneses, P . R ., and D utea, L . V .,
1988, Spectral and spatial attribute evaluation of SPOT data in geological mapping
of Precambrian terrains in semi-arid environment of Brazil. SPOT 1: Image utilization,
assessment, results, Proceedings CNES Conference, Paris, France, Cepadue s Editions
( Paris: CNES), pp. 851± 860.
P ellemans, A . H . J . M ., J ordans, R . W . L ., and A llewijn, R ., 1993, Merging multispectral
and panchromatic SPOT images with respect to the radiometric properties of the
sensor. Photogrammetric Engineering and Remote Sensing, 59, 81± 87.
P erlant, F ., 1992, Example of SPOT/ ERS-1 complementary. Space at the Service of our
Environment, Proceedings First ERS-1 Symposium, 4± 6 November 1992, Cannes, France
( Paris: ESA), pp. 563± 568.
P erlant, F ., S empere, J .-P ., and G uerre, L .-F ., 1994, Production of SPOT/ ERSI image
maps. Proceedings of First ERS-1 Pilot Project Workshop, T oledo, Spain, 22± 24 June
1994 , SP-365 ( Paris: European Space Agency), pp. 331± 336.
P ohl, C ., 1995, SPOT/ ERS Image maps: Topographic map updating in Indonesia. SPOT
MAGAZINE, December 1995, 17± 19.
P ohl, C ., 1996, Geometric aspects of multisensor image fusion for topographic map updating
in the humid Tropics. IT C publication No. 39 ( Enschede: ITC), ISBN 90 6164 121 7.
P ohl, C ., and G enderen, J . L . van, 1993, Geometric integration of multi-image information.
Space at the Service of our Environment, Proceedings of the Second ERS-1 Symposium,
11± 14 October 1993, Hamburg, Germany, ESA SP-361 (Paris: European Space Agency),
pp. 1255± 1260.
P ohl, C ., and G enderen, J . L . van, 1995, Image fusion of microwave and optical remote
sensing data for map updating in the Tropics. Image and Signal Processing for Remote
Sensing, Proceedings EUROPT O ’95, Paris, France, 25± 29 September 1995 , SPIE
Vol. 2579, pp. 2± 10.
P ohl, C ., W ang, Y ., and K oopmans, B . N ., 1994. The 1995 ¯ ood in the Netherlands from
space. IT C Journal , 1994-4, 414± 415.
P ohl, C ., K oopmans, B . N ., and W ang, Y ., 1995, Provision of ¯ ood information in the
Netherlands. In Application Achievements of ERS-1, New V iews of the Earth,
Preliminary V ersion (Paris: European Space Agency), pp. 122± 123.
P olidori, L ., and M angolini, M ., 1996, Potentials and limitations of multisensor data fusion.
Fusion of Earth Data, Proceedings EARSeL Conference, Cannes, France, 6± 8 February
1996 , pp. 13± 19.
P radines, D ., 1986 , Improving SPOT image size and multispectral resolution. Earth Remote
Sensing using the L andsat T hematic Mapper and SPOT Systems, Proceedings SPIE
Conference, 660, pp. 78± 102.
P rice, J . C ., 1987, Combining panchromatic and multispectral imagery from dual resolution
satellite instruments. Remote Sensing of Environment , 21, 119± 128.
R aggam, J ., A lmer, A ., and S trobl, D ., 1994, A combination of SAR and optical line scanner
imagery for stereoscopic extraction of 3-D data. ISPRS Journal of Photogrammetry
and Remote Sensing, 49, 11± 21.
852 C. Pohl and J. L . van Genderen

R amseier, R . O ., E mmons, A ., A rmour, B ., and G arrity, C ., 1993, Fusion of ERS-1 SAR

and SSM / I Ice data. Space at the Service of our Environment, Proceedings of the Second
ERS-1 Symposium, 11± 14 October 1993, Hamburg, Germany, ESA SP-361 ( Paris:
European Space Agency), pp. 361± 368.
R anchin, T ., and W ald, L ., 1993, The wavelet transform for the analysis of remotely sensed
images. International Journal of Remote Sensing, 14, 615± 619.
R anchin, T ., W ald, L ., and M angolini, M ., 1996, The ARSIS method: a general solution
for improving spatial resolution of images by the means of sensor fusion. Fusion of
Earth Data, Proceedings EARSeL Conference, Cannes, France, 6± 8 February 1996
( Paris: European Space Agency).
R ast, M ., J askolla, M ., and A ranson, F . K ., 1991, Comparative digital analysis of Seasat-
SAR and Landsat-TM data for Iceland. International Journal of Remote Sensing,
12, 527± 544.
R ay, P . K . C . R oy, A . K ., and P rabhakaran, B ., 1995, Evaluation and integration of ERS-1
SAR and optical sensor data (TM and IRS) for geological investigations.
Photonirvachak, Journal of the Indian Society of Remote Sensing, 23, 77± 86.
R ebillard, P ., and N guyen, P . T ., 1982, An exploration of co-registered SIR-A, SEASAT
and Landsat images. RS of Environment, RS for Exploration Geology, Proceedings
International Symposium, Second T hematic Conference, Forth W orth, U.S.A. (Forth
Worth: ISRS), pp. 109± 118.
R eimchen, T . H . F ., 1982, Location of economic gypsum deposits by integration of multi-
spectral, multitemporal, geophysical, -chemical, -botanical and geological data. RS of
Environment, RS for Exploration Geology, Proceedings International Symposium, Second
T hematic Conference, Forth Worth, U.S.A . (Forth Worth: ISRS), pp. 119± 125.
R ichards, J . A ., 1984, Thematic mapping from multitemporal image data using the principal
component transformation. Remote Sensing of Environment , 16, 35± 46.
R ogers, R . H ., and W ood, L ., 1990, The history and status of merging multiple sensor data:
an overview. T echnical Papers 1990, ACSM-ASPRS Annual Convention, Image
Processing and Remote Sensing, 4, 352± 360.
R othery, D . A ., and F rancis, P . W ., 1987, Synergistic use of MOMS-01 and Landsat TM
data. International Journal of Remote Sensing, 8, 501± 508.
R uss, J . C ., 1995, T he image processing handbook . Second edn (London: CRC Press).
S chistad-S olberg, A . H ., J ain, A . K ., and T axt, T ., 1994, Multisource classi® cation of
remotely sensed data: fusion of Landsat TM and SAR images. I.E.E.E. T ransactions
on Geoscience and Remote Sensing, 32, 768± 778.
S harpe, B ., and K err, A ., 1991, Multichannel fusion technique for resolution enhancement
during ground processing. Proceedings of the I.E.E.E. International Geoscience and
Remote Sensing Symposium (IGARSS ’91), Espoo, Finland, 3± 6 June 1991 (New York:
I.E.E.E.), pp. 1459± 1462.
S heffield, C ., 1985, Selecting band combinations from multispectral data. Photogrammetric
Engineering and Remote Sensing, 51, 681± 687.
S hen, S . S ., 1990, Summary of types of data fusion methods utilized in workshop papers.
Multisource Data Integration in Remote Sensing, Proceedings of W orkshop, Maryland,
U.S.A., 14± 15 June 1990 , NASA Conference Publication 3099 (Greenbelt, MD: NASA ),
pp. 145± 149.
S hettigara, V . K ., 1992, A generalized component substitution technique for spatial enhance-
ment of multispectral images using a higher resolution data set. Photogrammetric
Engineering and Remote Sensing, 58, 561± 567.
S hufelt, J ., and M c K eown, D . M ., 1990, Use of information fusion to improve the detection
of man-made structures in aerial imagery. Multisource Data Integration in Remote
Sensing, Proceedings of W orkshop, Maryland, U.S.A., 14± 15 June 1990 , NASA
Conference Publication 3099 (Greenbelt, MD: NASA), pp. 94± 110.
S imard, R ., 1982, Improved spatial and altimetric information from SPOT composite imagery.
Proceedings ISPRS Conference, Forth W orth, U.S.A., 6± 10 December 1982 (Forth
Worth: International Archive of Photogrammetry and Remote Sensing ), pp. 433± 440.
S ingh, A ., 1989, Digital change detection techniques using remotely-sensed data. International
Journal of Remote Sensing, 10, 989± 1003.
S ingh, A ., and H arrison, A ., 1985, Standardized principal components. International Journal
of Remote Sensing, 6, 883± 396.
 Multisensor image f usion in remote sensing 853

S mara, Y ., B elhadj- A issa, A ., and S ansal, B ., 1996, Decorrelation and IHS color transforms
applied for multisource satellite images integration. Fusion of Earth Data, Proceedings
EARSeL Conference, Cannes, France, 6± 8 February 1996 ( Paris: European Space
Agency), pp. 107± 112.
S trobl, D ., R aggam, J ., and B uchroithner, M . F ., 1990, Terrain correction geocoding of a
multi-sensor image data set. Proceedings 10th EARSeL Symposium, T oulouse, France
( Paris: European Space Agency), pp. 98± 107.
S uits, G ., M alila, W ., and W eller, T ., 1988, Procedures for using signals from one sensor
as substitutes for signals of another. Remote Sensing of Environment , 25, 395± 408.
T anaka, S ., S ugimura, T ., and H igashi, M ., 1989, High resolution satellite image map from
the SPOT and Landsat TM data. Advanced Space Research, 9, 115± 120.
T aranik, J . V ., 1988, First results of international investigations of the applciations of SPOT-1
data to geologic problems, mineral and energy exploration. SPOT 1: Image utilization,
assessment, results, Proceedings CNES Conference, Paris, France, Cepadue s Editions
( Paris: CNES), pp. 701± 708.
T auch, R ., and K ähler, M ., 1988, Improving the quality of satellite image maps by various
processing techniques. International Archives of Photogrammetry and Remote Sensing,
Proceedings XV I ISPRS Congress, Kyoto, Japan (Kyoto: International Archives of
Photogrammetry and Remote Sensing), pp. IV238± IV247.
T oll, D . L ., 1985, Analysis of digital Landsat MSS and Seasat SAR data for use in
discriminating land cover at the urban fringe of Denver, Colorado. International
Journal of Remote Sensing, 6, 1209± 1229.
T outin, T ., 1994, Multisource data integration with an integrated and uni® ed geometric
modelling. Proceedings of 14th EARSeL Symposium on Sensors and Environmental
Applications, 6± 8 June 1994, GoÈ teborg, Sweden (Paris: European Space Agency),
pp. 163± 174.
T outin, T ., and R ivard, B ., 1995, A new tool for depth perception of multi-source data.
Photogrammetric Engineering and Remote Sensing, 61, 1209± 1211.
U laby, F . T ., L i, R . Y ., and S hanmugan, K . S ., 1982, Crop classi® cation using airborne
radar and Landsat data. I.E.E.E. T ransactions in Geoscience and Remote Sensing,
20, 42± 51.
V ornberger, P . L ., and B lindschadler, R . A ., 1992, Multi-spectral analysis of ice sheets
using co-registered SAR and TM imagery. International Journal of Remote Sensing,
13, 637± 645.
W ang, Y ., K oopmans, B . N ., and P ohl, C ., 1995, The 1995 ¯ ood in The Netherlands
monitored from space Ð a multi-sensor approach. International Journal of Remote
Sensing, 16, 2735± 2739.
W elch, R ., and E hlers, M ., 1987, Merging multiresolution SPOT HRV and Landsat TM
data. Photogrammetric Engineering and Remote Sensing, 53, 301± 303.
W elch, R ., and E hlers, M ., 1988, Cartographic feature extraction from integrated SIR-B
and Landsat TM images. International Journal of Remote Sensing, 9, 873± 889.
W elch, R ., J ordon, T . R ., and E hlers, M ., 1985, Comparative evaluations of geodetic
accuracy and cartographic potential of Landsat-4/ 5 TM image data. Photogrammetric
Engineering and Remote Sensing, 51, 1249± 1262.
W elch, R ., J ordon, T . R ., and L uvall, J . C ., 1990, Geocoding and stereo display of tropical
forest multisensor data sets. Photogrammetric Engineering and Remote Sensing, 56,
1389± 1392.
W eydahl, D . J ., 1992, Temporal change detection in ERS-1 SAR images. International
Space Year: Space Remote Sensing, Proceedings 12th Annual I.E.E.E. International
Geoscience and Remote Sensing Symposium (IGARSS ’92), Houston, U.S.A. , edited by
R. Williamson with collaboration of T. Stein (New York: I.E.E.E.), pp. 1346± 1348.
W eydahl, D . J ., 1993 , Multitemporal analysis of ERS-1 SAR images over land areas. Better
Understanding of Earth Environment, Proceedings 13th Annual I.E.E.E. International
Geoscience and Remote Sensing Symposium (IGARSS ’93), T okyo, Japan, 18± 21 August
1993 (New York: I.E.E.E.), pp. 1459± 1461.
W ilkinson, G . G ., F olving, S ., K anellopoulos, I ., M c C ormick, N ., F ullerton, K ., and
M e’ gier, J ., 1995, Forest mapping from multi-source satellite data using neural network
classi® ers Ð An experiment in Portugal. Remote Sensing Reviews, 12, 83± 106.
854 Multisensor image f usion in remote sensing

Y e’sou, H ., B esnus, Y ., and R olet, J ., 1993 a, Extraction of spectral information from Landsat
TM data and merger with SPOT panchromatic imageryÐ a contribution to the study
of geological structures. ISPRS Journal of Photogrammetry and Remote Sensing, 48,
23± 36.
Y e’sou, H ., B esnus, Y ., R olet, J ., and P ion, J . C ., 1993 b, Merging Seasat and SPOT imagery
for the study of geologic structures in a temperate agricultural region. Remote Sensing
of Environment , 43, 265± 280.
Y e’sou, H ., B esnus, Y ., and R olet, J ., 1994, Perception of a geological body using multiple
source remotely-sensed dataÐ relative in¯ uence of the spectral content and the spatial
resolution. International Journal of Remote Sensing, 15, 2495± 2510.
Y ildimi, H ., A lparstan, E ., and O zel, M . E ., 1996, Data fusion of sharpened di € erence
image by IHS transform. Fusion of Earth Data, Proceedings EARSeL Conference,
Cannes, France, 6± 8 February 1996 ( Paris: European Space Agency), pp. 125± 130.
Z obrist, A . L ., B lackwell, R . J ., and S tromberg, W . D ., 1979, Integration of Landsat,
Seasat, and other geo-data sources. Proceedings 13th International Symposium on
Remote Sensing of Environment, ERIM, Ann Arbor, U.S.A., 23± 27 April 1979 (Ann
Arbor: E.R.I.M.), pp. 271± 280.