State-Of-The-Art of Geometric Correction of Remote Sensing Data: A Data Fusion Perspective

International Journal of Image and Data Fusion
Vol. 2, No. 1, March 2011, 3–35
State-of-the-art of geometric correction of remote sensing

data: a data fusion perspective
Thierry Toutin*
Canada Centre for Remote Sensing, Natural Resources Canada,

588 Booth Street, Ottawa, ON K1A 0Y7, Canada
(Received 23 July 2010; final version received 5 November 2010)
The three-dimensional (3D) geometric processing and ortho-rectification of

remote sensing (RS) data becomes a key issue as being the first step in multi-
source multi-format data fusion in geographic information systems. The fusion of
image data (visible and microwave, panchromatic and multi bands, polarimetric
bands, passive and active, etc.), their metadata (GPS, star trackers, inertial
system, lens and focal plane, etc.), the associated 3D cartographic data (ground
control points, contour lines, digital terrain model (DEM), planimetric features,
etc.) is thus a requisite to perform first 3D precise geometric correction and then
the ortho-rectification process with DEM. This article presents an update of the
state-of-the-art of geometric correction with the source of geometric distortions,
the different mathematical models, the methods, algorithms and processing
steps to track finally the error propagation during the fusion of the different RS
and cartographic data from the image acquisition to the ortho-rectification
processes.
Keywords: multi-format RS/GIS data fusion; geometric distortions and
correction; mathematical modelling; error propagation
1. Introduction
Raw remote sensing (RS) data contain such significant geometric distortions that they
cannot be fused with any other data (RS, topographic, cartographic, semantic, etc.) in a
geographic information system (GIS). Being the first step of multi-source multi-format
data fusion for geomatics applications geometric and radiometric processing must be
adapted to the nature and characteristics of each RS data in order to keep the best
information in the fused products. Each raw data must be separately, or preferably
simultaneously, converted into an ortho-image so that each component of the data set can
be registered, compared, fused, etc. in a GIS. Consequently, the ortho-rectification process
is a large data fusion process, which included data from the platform, the sensor, the
Earth, the map projection and the GIS.
One must admit that since 1970s, a drastic evolution of on-board systems with large
scientific and technology progresses introduced new needs and requirements for
data fusion, such as these non-exhaustive examples: increased number and enhanced
*Email: thierry.toutin@ccrs.nrcan.gc.ca
ISSN 1947–9832 print/ISSN 1947–9824 online

ß 2011 Her Majesty the Queen in Right of Canada
DOI: 10.1080/19479832.2010.539188
http://www.informaworld.com
4 T. Toutin
electro-optical (EO) sensors resulting in more and better information extraction;

development of performing non-imaging instruments, such as inertial navigation systems
(INS), star trackers and orbit positioning resulting in more accurate metadata, three-
dimensional (3D) digital data and representation with the development of advanced digital
mapping and information technology systems requiring new methods and processing
algorithms. The geometric correction is thus no more a simple independent process in the
data fusion but is an integral part of this full process where all the previously mentioned
issues and data have to be simultaneously taken into account to use the full potential of
these new capabilities. While the radiometric issues of the multi-source RS data fusion was
already well-reviewed (Lu et al. 2010, Zhang 2010), the geometric issues were rarely
addressed, only for some specific aspects (Palubinskas et al. 2010, Suri et al. 2010, Toutin
et al. 2010). It is thus important to update in a global perspective the problems and the
solutions recently adopted in the multi-source data fusion to correct RS images
geometrically with their fusion with other multi-source and – format data. This state-of-
the-art paper will then address with a data fusion perspective the latest developments and
research studies from around the world related to:
. the source of geometric distortions and deformations with a taxonomy as well as
the fusion of correlated distortions (Section 2);
. the modelling of the distortions with different two-dimensional (2D)/3D physical/
empirical models and mathematical functions as well as the fusion of some
distortion parameters and their associated mathematical functions (Section 3) and
. the method and algorithms of 3D geometric correction with the fusion of the used
EO/GIS data (Section 4).
Comparisons between the models and mathematical functions, their applicability and
their performance on different types of images (frame camera, visible and infra-red (VIR)
oscillating or push-broom scanners; side looking antenna radar (SLAR) or synthetic
aperture radar (SAR) sensors; high, medium or low resolution) are addressed. The errors
with their propagation from the input data to the final results are also evaluated through
the full processing steps.
2. Taxonomy of geometric distortions

Each EO image acquisition system produces unique geometric distortions in its raw images
and consequently the geometry of these images in its own local coordinate system does not
correspond to other images, to the terrain and to user’s specific map projection. Obviously,
the geometric distortions vary considerably with different factors such as the platform
(airborne and satellite), the sensor (VIR and SAR; total field of view (FOV) or beam angle,
low to high resolution), the associated scanner (whiskbroom, pushbroom, frame, etc.) but
also the application for the data fusion. It is, however, possible to make taxonomy of these
distortions.
The source of distortions (Table 1) can be grouped into two broad categories: the
Observer or the acquisition system (platform, imaging sensor and other measuring
instruments, such as gyroscope, stellar sensors, etc.) and the Observed (atmosphere and
Earth). In addition to these distortions, the deformations related to the map projection
International Journal of Image and Data Fusion 5
Table 1. Description of error sources for the two categories, the Observer and the Observed, with the
different sub-categories.
Category Sub-Category Description of Error Sources

Platform Variation of the movement
The Observer (spaceborne or Variation in platform attitude (low to high
airborne) frequencies)
or Variation in sensor mechanics (scan rate,
Sensor scanning velocity, etc.)
The (VIR, SAR or HR) Lens distortions, viewing/look angles
Acquisition Panoramic effect with the field of view
System Measuring Time-variations or drift
instruments Clock synchronicity
Atmosphere Refraction and turbulence
The Observed Earth Curvature, rotation, topographic effect
Map G e oi d t o el l ips o i d
Ellipsoid to map
have to be taken into account because the terrain and most of GIS end-user applications
are generally represented and performed, respectively, in a topographic map space and not
in a referenced ellipsoid. The map deformations are logically included in the distortions
of the Observed.
Previous studies made a second-level categorisation into low, medium and high-
frequency distortions (Friedmann et al. 1983), where ‘frequency’ is determined or
compared to the image acquisition time. Examples of low-, medium- and high-frequency
distortions are orbit variations, Earth rotation and local topographic effects, respectively.
While this categorisation was suitable in the 1980s when there were very few RS and digital
systems, today with so many different acquisition systems, it is no longer acceptable
because it differs with each acquisition system. For example, attitude variations are a high-
frequency distortion for QuickBird or airborne push-broom scanner, a medium-frequency
distortion for SPOT-HRV and Landsat-ETMþ, a low-frequency distortion for Landsat-
MSS but no distortions for MERIS. As examples, an assessment of the influences of
satellite viewing angle, Earth curvature, relief height, geographical location and digital
terrain model (DEM) characteristics on planimetric displacement was performed on
high-resolution (HR) satellite imagery (Zhang et al. 2001).
The geometric distortions and their error sources of Table 1 are predictable and
generally well-understood. Some of these distortions, especially those related to the
instrumentation, are systematic and generally corrected at ground receiving stations or by
the image vendors. Others, for example those related to the atmosphere, are not taken into
account and corrected because they are specific to each acquisition time and location while
6 T. Toutin
information on the atmosphere is rarely available. They also are geometrically negligible
for low-to-medium resolution images.
2.1 Distortions related to the platform

Some basics on satellite orbits and celestial mechanics are useful to better understand the
platform-related distortions. EO satellites obey the celestial mechanical laws as defined by
Newton and Kepler for an unperturbed trajectory (Keplerian orbit) and by Gauss and
Lagrange for a perturbed trajectory (osculatory orbit; Escobal 1965, Centre National
d’Études Spatiales 1980). A number of perturbations (Earth gravity and surface
irregularities, atmospheric drag, etc.) slowly change the Keplerian orbit based on the
two-body attraction of Newton’s law into an osculatory orbit (Centre National d’Études
Spatiales 1980).
Most, if not all, of the civilian EO spacecrafts have near-Earth, retrograde, quasi-
circular, quasi-polar, geosynchronous and sunsynchronous orbits. Near-Earth orbits
(more than 300 km altitude) are high enough to reduce the atmospheric drag. Retrograde
orbits with 90–180 inclination are westward-launched orbits, which require extra fuel to
compensate for the Earth’s rotation, but they provide the only solution for obtaining
sunsynchronous orbits. Quasi-circular orbits avoiding large changes in altitude enable
images with similar scales to be acquired, which is desirable for EO. Quasi-polar orbits
with 90–100 inclination enable sensors to image the entire Earth, including most of the
poles. Because geosynchronous orbits have a repeating ground track, they have an orbital
period that is an integer multiple of the Earth’s sidereal rotation period. This integer
multiple is called the repeat cycle. The special case of a geosynchronous orbit that is
circular and directly above the Equator is called a geostationary orbit. The satellite track
on the ground, also called path, is kept fixed within certain limits related to orbit
maintenance accuracy. Paths are artificially divided into squared scenes at regular
intervals, generating rows. The main advantage is that the satellite follows a fixed pattern
on the Earth, which is desirable for operational EO systems. Sunsynchronous orbits enable
satellites to pass overhead at the same local solar time and thus to acquire images with
almost-identical illumination lighting conditions. While the comparison of multi-date
images is easier with a sunsynchronous orbit, this approach has some disadvantages,
especially as variations of illumination reveal different structural details. In addition,
sunsynchronous orbits require retrograde orbits and strict relationships between orbital
parameters (mainly inclination and height), which must be preserved during the satellite’s
lifetime.
The distortions associated with the near-Earth orbit satellite are primarily due to
Newton’s laws interaction the platform and Earth (Earth’s gravity, shape and movement
generating a quasi-elliptic movement; Escobal 1965, Centre National d’Études Spatiales
1980, Light et al. 1980). The satellite has six degrees of freedom in space, which can be
determined either by (Toutin 1983, Kim and Dowman 2006):
. its position related to the Earth’s centre (Figure 1, left): three geo-positioning
parameters (X, Y, Z) and orientation (a solid angle);
. its position/velocity (Figure 1, right): the attitude (roll, pitch, yaw) a specific kind
of Euler angles or
Figure 1. Orbital (left) and attitude (right) parameters defining the orbit and the satellite on its orbit
as a function of time. Courtesy and copyright Serge Riazanoff, VisioTerra, 2009.
. its osculatory orbit (Figure 2): six parameters related to the instantaneous orbit
(semi-major axis, eccentricity, inclination, ascending-node longitude, perigee
argument, mean anomaly).
Depending of the acquisition time (used as 7th parameter) and the size of the image, all
these parameters are time-dependent due to the orbital perturbations and thus generate
a range of non-systematic distortions. All these are non-systematic and time-dependent
distortions are not predictable and must be evaluated for each image from satellite
tracking data and/or ground-control information. Some effects of these distortions
include:
. The platform altitude variation (Z-axis) is the most critical among the
position parameters. It can change the pixel spacing in the across-track direction
while X- and Y-parameters generate only a translation in the sensed ground area.
The SAR platform altitude will alter the SAR incidence angles and thus the
viewing geometry. In addition, X- and Z- variations of the platform will generate
some differential shifts in the range measurement between lines. However, the
altitude variation of the satellite (around 8–10 m/s for near-Earth orbit at around
800 km altitude) is not too significant over the time required to acquire a full scene
(5–10 s for high to medium resolution satellite sensors): it only represents 1/8000
relative pixel variation. For SPOT5 panchromatic super mode (12,000 pixels with
2.5 m spacing), it generates 0.3 mm variation in pixel spacing, which cumulatively
generates around 2 m error (0.0003 m per pixel multiplied by 6000 pixels) at the
edge of the image.
. The platform velocity variations can change the line spacing or create line gaps/
overlaps for optical data and Doppler variations for SAR data. Variations in
spacecraft velocity only cause distortion in the along-track direction.
8 T. Toutin
Figure 2. Description of a satellite osculatory orbit and its approximation by a Keplerian ellipse.
X, Y, Z are the positions in a geocentric frame reference system. I is the orbit inclination;
is the
longitude of the ascending node (N); ! is the argument of the perigee (P); (! þ v) is the argument
of the satellite and is the distance from the Earth centre (O) and the satellite.
. The platform attitude variation (Figure 3) can change the orientation and the
shape of VIR images, but it does not affect SAR image geometry. The roll is the
rotation around the flight vector (X-axis, Figure 3, right), hence in a ‘wing down’
direction, and its variation causes lateral shifts and scale change in the across-
track direction. The pitch is the vertical rotation of the platform, in the ‘nose up’
plane and its variation results in scan-line spacing changes. The yaw is the
rotation around the vertical axis and its variation changes the orientation of each
scanned line resulting in a skew between the scan lines. The variations in the
platform attitude can be severe in term of location on the ground (absolute offset
of hundreds of metres) over the time required to scan a full scene (5–10 s) and are
significant when the variation are sudden over few lines (relative offset of tens of
metres within milliseconds).
2.2 Distortions related to the sensor

The remaining sensor-related distortions include:
. The calibration parameters uncertainty, such as for the lens distortions (focal
length, principal point, detector position, decentring, radial/tangential, etc.) and
Figure 3. Example of SPOT attitude variations (left part) and its impact on image geometry (right
part): pitch (left), roll (centre) and yaw (right).
the instantaneous field of view (IFOV) for VIR sensors. The range gate delay
(timing) for SAR sensors is corrected in the SAR processor. The uncertainties in
the focal length change the pixel size and those in the principal point a systematic
shift of the scan lines and of the full image. The lens distortion causes imaged
positions to be distorted along radial/tangential lines from the principal point.
This distortion, which depends of the detector position in the focal plane, has
been corrected for HR sensors (below 10-m resolution) and for the pre-processed/
geo-referenced images. Information is sometimes included in the metadata
(e.g. SPOT5).
. The panoramic distortion in combination with the Earth curvature and the
topographic relief changes the ground pixel sampling along the swath for optical
and SAR data. Panoramic distortion occurs for large swath sensors (MERIS,
MODIS, ScanSAR, etc.). As the sensor scans across each line, the distance from
the sensor to the ground increases further away from the centre of the swath and
scans a larger area as it moves closer to the edges. The further away from nadir is
an object, the larger is the compression. This effect results in the compression of
image features at points away from the nadir and is called tangential scale
distortion. However, it is the reverse for SAR: the compression of image features
increases from far to near edges because the ground range resolution is reduced at
high incidence angles. As example, when the SAR resolution is 10 m, the ground
range resolution will be 29.0 and 10.1 m at 10 and 70 incidence angle,
respectively. This compression still remains for images in the slant-range geometry
but is taken into account in the SAR processor when images are generated in the
ground-range geometry.
The main distortions related to the non-imaging sensors can be caused by the
clock synchronicity between all instruments and the variations or drift of the time for
an instrument. Because the time is the only parameter making a link between all
instruments, it can produce different errors in their data fusion depending of the measuring
instrument. Generally, the corrections, if possible, will be performed at ground
receiving stations.
10 T. Toutin
2.3 Distortions related to the Earth

The remaining Earth-related distortions include:
. Its rotation generates latitude-dependent displacements from the east to the
west between scanned lines. The eastward rotation of the Earth, during a satellite
orbit, causes the sweep of scanning systems to cover an area slightly to the west
of each previous scan and which cumulates as the number of scan lines
accumulates. The resultant imagery is thus westward skewed across the image.
The amount of linear rotation varies with the scanning time: longer for Landsat
than for SPOT. This is known as skew distortion. For SAR images, the rotation
has an impact on the Doppler centroid but it is taken into account in the SAR
processor.
. Its curvature creates variation of the ground spacing in the along-track direction,
which increases with the distance from the ground nadir point. Its effect is related
to both the off-nadir viewing angle and the FOV, which increase the distance from
the nadir point. For SAR images, the curvature is taken into account and
corrected in the SAR processor.
. Its topographic relief generates a displacement in the viewing direction. This
displacement is away from nadir for optical data and to the nadir for SAR data
generating a compression known as foreshortening. When the topography is
severe with strong slopes, the foreshortening is transformed into layover (a kind
of inverse relief), because the top of the mountain is imaged before its bottom
(Schreier 1993). This effect is more severe at small incidence angles. On the other
hand, the radar shadow (or occluded areas) due to no radar return signal in steep
terrain is more severe at large incidence angles.
As the EO satellite data and the extracted information are afterwards fused with multi-
source multi-format data, generally in a GIS, they have to be finally projected into the
user’s map reference system. The remaining deformations associated with the map
projection are thus:
. the approximation of the geoid by a reference ellipsoid; and
. the projection of the reference ellipsoid on a plane surface (generally a conformal
projection, preserving angles).
However, some distortions of the platform, the imaging sensor, the non-imaging
measuring instruments, the Earth and the cartographic projection are correlated because
they produce the same impact on the ground. Consequently taking into account the full
geometry of viewing together instead of separately for each element, the fusion of the
correlated distortions can be performed into different combined distortions. This fusion of
distortions is theoretically better than considering each distortion separately, avoiding
over-parametrisation. Some non-exhaustive examples of distortions fusion include:
. the distortions due to the velocity, the altitude and the pitch of the platform, the
detection signal time of the sensor and the Earth rotation component in the
along-track direction;
. the distortions due to the platform roll, the across-track angle of the sensor,
and the Earth curvature and
. the distortions due the linear time-dependent variation of the platform altitude,
the linear time-dependent roll variation and the Earth rotation, etc.
3. Geometric modelling of distortions

All these remaining geometric distortions require geometric models and mathematical
functions to perform the geometric correction of imagery: either through empirical models
(such as polynomial or rational functions, RFs) or with rigorous physical and
deterministic models.
3.1 2D/3D empirical models

The empirical models can be used when the parameters of the acquisition systems or a
rigorous 3D physical model are not available. Since they do not reflect the source of
distortions described previously, these models do not require a priori information on any
component of the total system (platform, sensor, Earth and map projection).
These empirical models are based on different mathematical functions:
(1) 2D polynomial functions, such as:
m X
X n
P2D ðXYÞ ¼ aij Xi Yj ð1Þ
i¼0 j¼0
(2) 3D polynomial functions such as:

m X
X n X
p
P3D ðXYZÞ ¼ aijk Xi Yj Zk ð2Þ
i¼0 j¼0
(3) 3D RFs, such as:

Pm Pn Pp
i¼0 j¼0 k¼0 aijk Xi Yj Zk
R3D ðXYZÞ ¼ Pm Pn Pp , ð3Þ
i¼0 j¼0 k¼0 bijk Xi Yj Zk
where X, Y, Z are the terrain or cartographic coordinates; i, j, k are the integer increments;
and m, n and p are integer values, generally comprised between 0 and 3 equal and being the
degree of the polynomial functions.
Each 2D 1st-, 2nd- and 3rd-degree polynomial functions will then have 3-, 6- and
10-term unknowns. Each 3D 1st-, 2nd- and 3rd-degree polynomial functions will then have
4, 10 and 20-term unknowns. The 1st-degree polynomial functions are also called affine
transformations. Each 3D 1st-, 2nd- and 3rd-degree RFs will have 8-, 20- and 40-term
unknowns. In fact, the 3D RFs are extensions of the collinearity equations (Section 3.3),
which are equivalent to 1st-degree RFs. Depending of the imaging geometry in each axis
(flight and scan), the degree of polynomial functions (numerator and denominator for
RFs) can be different and/or specific terms, such as XZ for 2D or XZ, YZ2 or Z3, etc.
for 3D, can be differently ignored of the polynomial functions, when these terms cannot be
related to any physical element of the image acquisition geometry. These ‘intelligent’
polynomial functions reflect then better the geometry in both axes and reduce the over-
parametrisation and the correlation between terms. Okamoto (1981, 1988) already applied
this reduction of terms for one-dimensional (1D) central perspective photographs and line-
scanners, respectively. Since the simplest solution, the 2D polynomial functions applied
1970–1980s are fortunately no more used with the new EO sensors, only 3D functions are
now discussed.
12 T. Toutin
3.1.1 3D polynomial functions

The 3D polynomial functions are an extension of the 2D polynomial function by adding
Z-terms related to the third dimension of the terrain. However, they are prone to the same
problems as any empirical functions, except for the relief: i.e. they are applicable to small
images, they need numerous ground control points (GCPs) regularly distributed, they
correct locally at GCPs, they are very sensitive to errors and they have a lack of robustness
and consistencies in operational environment. Their use should thus be limited to small
images or to systematically corrected images, where all distortions except the relief have
been corrected. For these reasons, 2nd-degree conformal polynomial functions have been
primarily used in aerial photogrammetry during the 1960s (Baetslé 1966, Schut 1966). Due
to the larger size of satellite images, they have been mainly used with geo-referenced
images: SPOT-HRV (level 1 and 2) using 1st-degree functions (Baltsavias and Stallmann
1992, Okamoto et al. 1998); SPOT-HRV (level 1B) and Landsat-TM (level bulk or geo-
referenced) using 2nd-degree functions (Palà and Pons 1995). More recently, 1st-degree
affine functions were applied to IKONOS Geo-products (Ahn et al. 2001, Fraser et al.
2002a,b, Vassilopoulou et al. 2002). The terms related to terrain elevation in the 3D
polynomial function could be reduced to the linear term aiZ for VIR images and to the
quadratic terms aiZ and ajZ2 for SAR images whatever the degree of the polynomial
functions used. The main reasons are that the elevation distortion is linear and quadratic
for optical and SAR images, respectively, and there is no physical inter-relation in the
X and Z or Y and Z directions for most of the sensors used (e.g. with a zero-Doppler
SAR model).
In fact, Kratky (1989) already used 3rd or 4th degree 3D polynomial functions with
this term reduction to approximate its 3D physical model developed for SPOT raw images.
The main reason for his SPOT model was that the real-time computation for implementing
his specific physical model solution was not feasible on a stereo-workstation. He would
certainly not do this approximation with the higher performance computers now available.
More recently, tests were also performed using Kratky’s polynomial functions with
IKONOS Geo products acquired with near-nadir viewing over high relief areas
(Kersten et al. 2000, Vassilopoulou et al. 2002). The second study evaluated the
ortho-image errors over the GCPs and 1–2 m errors were achieved depending of their
number, definition and image-measurement accuracy, but this statistical evaluation is a
little biased by using checked data applied in the geometric correction process. However as
previously mentioned, the 3D affine transformation, also evaluated in the second study,
gave the same results than 4th-degree polynomial functions, but with much less GCPs.
3.1.2 3D rational functions

While occasionally used during the 1980s (Okamoto 1981, 1988), the interest in 3D RFs
has recently been renewed in the civilian photogrammetric and RS communities with the
launch of the first civilian HR IKONOS sensor in 1999 (Grodecki 2001, Fraser et al.
2002b, 2006, Fraser and Ravanbakhsh 2009, Tong et al. 2010 and more). Since sensor and
orbit parameters were not included in the metadata, 3D RFMs could be an alternative to
avoid the development of 3D physical models. The 3D RFMs can be used in two
approaches (Madani 1999):
(1) to approximate an already-solved existing 3D physical model (terrain-independent);
and
(2) to compute normally the unknowns of all the polynomial functions with GCPs
(terrain-dependent).
In some aspect, RFM can be considered as a ‘fused’ math model because each of
80 coefficients are not physically representative of a specific distortion but of combined
distortions, and can correct for part or more than one distortion.
3.1.2.1 Terrain-independent RFM approach. The first approach, inappropriately called

terrain-independent because the process still requests some GCPs to remove RFM bias
(see below), is performed in two steps. A 3D regular grid of the imaged terrain is first
defined and the image coordinates of the 3D grid ground points are computed using the
already-solved existing 3D physical model. These grid points and their 3D ground and 2D
image coordinates are then used as GCPs to resolve the 3D RFs and compute the
unknown terms of polynomial functions. Dowmann and Dolloff (2000) addressed the
advantages of this ‘universal real-time image geometry model’, such as universality,
confidentiality, efficiency and information transfer, as well as the disadvantages. In their
first development stage, the disadvantages can be summarised (Madani 1999):
. the inability to model local distortions (such as high-frequency variations);
. a limitation in the image size;
. the difficulty in the interpretation of the parameters due to the lack of physical
meaning;
. a potential failure to zero denominator and
. a potential correlation between the terms of polynomial functions.
Different strategy can be adopted to reduce these limitations. For the inability to
model local distortions, RFs should be applied to geo-referenced data with no systematic
distortions (such as IKONOS Geo images) rather than raw data with large geometric
distortions (such as QuickBird-2 or SPOT-5). For the limitation in the image size, the
image itself could thus to be subdivided into sub-images and separate 3D RFs are required
for each sub-image (Yang 2001). It results, however, in much more geometric but also
radiometric processing. For the potential failure and correlation, the non-significant
and/or high-correlated terms can be eliminated to avoid zero crossing and instability of
RFs depending of image geometry (Dowman and Dolloff 2000). To overcome some of
these problems, a ‘universal real-time image geometry model’ based on RFs has been
developed (OGC 1999). It is a dynamic RF of variable degrees, whose terms can be either
chosen as a function of the sensor geometry or eliminated using an iterative procedure
(Robertson 2003). This reduction of terms is then similar to the orientation theory
developed in the 1980s (Okamoto 1981, 1988). In addition, when the denominator
functions are omitted, RFs become thus simple 3D polynomial functions.
Image vendors and government agencies that do not want to deliver satellite/sensor
information with the image and commercial photogrammetric workstation suppliers are
the main users of this first approach. Image vendors thus provide with the image all the
parameters of 3D RFs. Consequently, the users can theoretically process the images
without GCP for generating ortho-images with DEM. This approach is adopted by
different image resellers around the world, which provide 3rd-degree RF parameters with
the small-FOV HR images: GeoEye with IKONOS and GeoEye-1 images (Grodecki
2001), DigitalGlobe and MacDonald, Dettwiler and Associates (MDA) with QuickBird-2
and WorldView-2 images (Hargreaves and Roberston 2001, Robertson 2003),
14 T. Toutin
the Indian Space Research Organization (ISRO) with Cartosat series and recently MDA
with Radarsat-2. This first approach was also tested under specific circumstances in an
academic environment with aerial photographs, and spaceborne VIR HR images (SPOT,
Eros-A, Formosat-2, Cartosat, QuickBird, etc.) by computing parameters of 1st- to 3rd-
degree RFs from already-solved 3D physical models (Tao and Hu 2001, Chen et al. 2006,
Bianconi et al. 2008), and recently with different spaceborne SAR images (Zhang et al.
2010). However, biases and random errors still exist after applying the RFs depending of
sensors and relief, and the results need to be post-processed with accurate user’s GCPs: (1)
the original RFs parameters can be refined with linear equations (Lee et al. 2002,
Fraser et al. 2006) or (2) the errors are corrected with 2D polynomial functions
(Fraser et al. 2002a,b, Tao and Hu 2002, Fraser and Hanley 2005, Wang et al. 2005,
Toutin 2006a, Tong et al. 2010 and more). The first solution achieves better results
than the second one.
3.1.2.2 Terrain-dependent RFM approach. The second approach, called terrain-

dependent, can be performed by the end-users with the same processing method as with
polynomial functions. Since there are 40 and 80 parameters for the four 2nd- and 3rd-
degree polynomial functions, a minimum of 20 and 40 GCPs, respectively, are required to
resolve the 3D RFs. However, the RFs do not model the physical reality of the image
acquisition geometry and they are sensitive to input errors, such as the 2D/3D polynomial
functions. Since RFs, such as the 3D polynomial functions, mainly correct locally at GCP
locations with some remaining distortions between GCPs (Tao and Hu 2002), many more
GCPs will be required to reduce their error propagation in an operational environment.
A piecewise approach as described previously (Yang 2001) should also be used for large
images (SPOT5, Landsat, IRS, Radarsat), which will increase the number of GCPs
proportionally to the number of sub-images, making the method inadequate in operational
environment. In summary, this RFM solution is highly dependent on the number,
accuracy and distribution of GCPs and the terrain relief (Tao and Hu 2002). Since this
approach is not efficient, it is not used any more.
3.2 Physical models

The physical models used to perform the geometric correction differ, depending on the
sensor, the platform and its image acquisition geometry (Figure 4):
. array camera systems (instantaneous acquisition), such as photogrammetric
cameras, Metric Camera (MC) or Large Format Camera (LFC);
. mechanical rotating or wiskbroom sensors, such as Landsat-MSS, TM or ETMþ;
. traditional push-broom scanners (CCD line acquisition with one to three optical
systems using synchronous scanning up to 30 in across- or along-track direction),
such as MERIS, SPOT-HRV/HRG/HRS, ASTER, ALOS-PRISM,
. new agile HR scanners (staggered/butted-CCD line acquisition with flexible
single-lens optical system using synchronous or asynchronous scanning up to 45
at 360 around their camera axes), such as IKONOS, QuickBird, WorldView and
. SAR sensors (side-looking acquisition), such as ENVISAT, Radarsat-1/2,
CosmoSkyMed, Terra-SAR-X.
Figure 4. Geometry of viewing of different types of scanners.
Although each sensor has its own unique characteristics, one can draw generalities for
the development of 2D/3D physical models in order to correct fully all distortions
described previously. The physical model should mathematically model all distortions of
the platform, the sensor and the Earth as well as the deformations of the cartographic
projection. The general starting points to derive the mathematical functions of the 3D
physical model are generally:
(1) the well-known collinearity condition and equations (Bonneval 1972, Wong 1980)
for VIR images:
m11 ðX X0 Þ þ m12 ðY Y0 Þ þ m13 ðZ Z0 Þ
x ¼ ðf Þ , ð4Þ

y ¼ ðf Þ , ð5Þ
where (x, y) are the image coordinates; (X, Y, Z) are the map coordinates;
(X0, Y0, Z0) are the projection centre coordinates; f is the focal length of the
VIR sensor and [mij] are the nine elements of the orthogonal 3-rotation matrix.
(2) The Doppler and range equations for radar images:
! ! ! !
2ð VS VP Þ ð S PÞ
f ¼ ! ! ð6Þ

S : P
! !

r ¼ S P ð7Þ
16 T. Toutin
! !
where f is the Doppler value; r is the range distance; S and VS are the sensor
! !
position and velocity; P and VP are the target point position and velocity on the
ground and is the radar wavelength.
The collinearity equations are valid for an instantaneous image or scanline acquisition,
such as photogrammetric cameras (LFC, MC), VIR scanner sensors (SPOT, Landsat) and
the Doppler-range equations are valid for a SAR scanline. However, since the parameters
of neighbouring scanlines of scanners are highly correlated along the orbit, it is possible to
link the exposure centres and rotation angles of the different scanlines to integrate
supplemental information (Poli 2007), such as:
. the ephemeris and attitude data using celestial mechanic laws for satellite
images; or
. the global positioning system (GPS) and INS data for airborne images.
Considerable research has been carried out to develop robust and rigorous 3D physical
models describing the acquisition geometry related to different types of images (VIR and
SAR, low, medium and high resolution) and of platforms (spaceborne and airborne):
. for low/medium-resolution VIR satellite images (Wong 1975, Bähr 1976, Masson
d’Autume 1979, Sawada et al. 1981, Khizhnichenko 1982, Friedmann et al. 1983,
Guichard 1983, Toutin 1983, Konecny et al. 1986, 1987, Salamonowicz 1986,
Gugan 1987, Kratky 1987, 1989, Shu 1987, Westin 1990, Kornus et al. 2000,
Sylvander et al. 2000, Westin 2000, Poli 2007);
. for HR VIR satellite images (Cheng and Toutin 1998, Radhadevi et al. 1998,
Hargreaves and Robertson 2001, Bouillon et al. 2002, 2006, Chen and Teo 2002,
Toutin 2003a, 2006b, Poli 2007);
. for SAR satellite images (Gracie et al. 1970, Leberl 1978, 1990, Wong et al. 1981,
Curlander 1982, Naraghi et al. 1983, Guindon and Adair 1992, Tannous and
Pikeroen 1994, Toutin 1995);
. for VIR airborne images (Derenyi and Konecny 1966, Konecny 1976, 1979, Ebner
and Muller 1986) and
. for airborne SLAR/SAR images (La Prade 1963, Gracie et al. 1970, Konecny
1970, Leberl 1972, 1990, Hoogeboom et al. 1984, Toutin 1995) and many others
afterwards.
However, the geometric correction process can address each distortion one-by-one and
step-by-step with a math function for each distortion, or simultaneously with the fusion of
the different math function into an integrated math model. This last solution is better and
more precise because it reduces the cumulative errors and their propagation. Both
solutions were developed with physical models: the ‘one-by-one’ and ‘step-by-step’
solutions with different math functions are generally applied at the ground receiving
station as independent processing boxes (platform, sensor, Earth and projection).
Processing a new orbit, a new sensor or a new image into a new ellipsoid and map
projection thus requires changing only ‘one distortion’ or ‘one step’ box with the new
characteristics and parameters. Conversely, the ‘fusion of the math function’ solution is
mainly applied by the users, with their own ‘fused’ math model, such as the collinearity
and the range-Doppler equations. However, the difficulty in developing the math model by
fusing all available data (orbit, attitude, sensor, focal lens, ground control, etc.) and their
associated math function could have forced some to consider the ‘step-by-step’ solution.
Nevertheless, to compute one ‘fused’ parameter is theoretically more precise than

computing each component, separately, avoiding correlation between parameter compo-
nents. Some examples of ‘fused’ parameters include:
. The total ‘orientation’ of the image is a fusion of the platform heading due to
orbital inclination, the linear time-variation of the platform roll and the Earth
rotation.
. The total ‘scale factor’ in across-track direction is a fusion of the pitch of the
platform, the viewing angle of the sensor perpendicular to the orbit and the
component of the Earth rotation in the across-track direction.
. The total ‘levelling angle’ in the along-track direction is a fusion of platform
pitch, the viewing angle of the sensor in the azimuth direction and the Earth
curvature; etc.
4. Methods, processing and errors

Whatever the mathematical models used, the geometric correction method and processing
steps are more and less the same. The processing steps, in which the data fusion
applied, are:
. acquisition of image(s) and pre-processing of sensor/image metadata (useless for
empirical models);
. acquisition of the ground data (control/check/pass) with 2D image coordinates
and 3D map coordinates;
. computation of the mathematical models, in which the different data are fused for
each image;
. image(s) rectification preferably with elevation data (contour lines, DEM).
4.1 Acquisition of images and metadata

With VIR images, different types of image data with different levels of pre-processing can
be obtained, but the different image providers unfortunately use a range of terminology to
name the same type of image data. Standardisation should be better defined, mainly for
the convenience of users:
. Raw images with only normalisation and calibration of the detectors (e.g. level 1A
for SPOT, EROS and Formosat, 1B1 for ALOS, Basic for QuickBird/
WorldView, Single Look Complex (SLC) for Radarsat) without any geometric
correction are satellite-track oriented. For SAR images, the rawest level
corresponds to the slant-range geometry. In addition, full metadata related to
sensor, satellite (ephemeris and attitude) and image are provided;
. geo-referenced images (e.g. level 1B for SPOT, 1SYS for Cartosat, Standard for
QuickBird/WorldView, 1G for Landsat-ETMþ, SGF/SGX for Radarsat)
corrected for systematic distortions due to the sensor, the platform and the
Earth rotation and curvature are satellite-track oriented. For SAR images, this
level corresponds to the ground-range geometry. Generally, the metadata related
to sensor and satellite are provided; some of metadata are related to the
1B processing; or
18 T. Toutin
. map-oriented images, also called geocoded images, (e.g. level 2A for SPOT, Geo
Standard for IKONOS, 1B2 for ALOS, SSG/SPG for Radarsat) corrected for the
same distortions as geo-referenced images are north oriented. For SAR images,
this level corresponds to the geocoded ellipsoid corrected geometry. Generally,
less metadata related to sensor and satellite are provided; most of metadata are
related to the 2A processing and the ellipsoid/map characteristics.
4.1.1 Raw ‘level 1A’ images

For the sake of understanding, the easiest terminology defined for SPOT images are used.
The raw ‘level 1A’ images are preferred by photogrammetrists because 3D physical models
derived from co-linearity equations are well-known and developed and easily used in
softcopy workstations. Since different 3D physical models are largely available for such
VIR images, raw 1A-type images should be now favoured by the RS community too.
Specific software to read and pre-process the appropriate metadata (ephemeris, attitude,
sensor and image characteristics) have to be realised for each image sensor according to
the 3D physical model used. Using celestial mechanics laws and Lagrange equations
(Escobal 1965, Centre National d’Études Spatiales 1980, Light et al. 1980) the ephemeris
(position and velocity, Figure 3, left) can be transformed into specific osculatory orbital
parameters (Figure 4) to reduce the time-related effects (Toutin 1983). Since the Lagrange
equations take into account the variation of the Earth gravitational potential to link the
different positions of the satellite during the image formation, it is more accurate and
robust than using a constant ellipse with 2nd-degree time-dependent polynomial functions
(Guichard 1983, Toutin 1983, Tannous and Pikeroen 1994, Bannari et al. 1995). This
statement is more applicable, regardless the sensor resolution, when paths or blocks
formed from ‘long-strip’ images from a same orbit are processed.
RFM 2nd approach was also applied with raw HR small-FOV images (QuickBird-2,
Formosat-2, etc.) since 2005 (Fraser and Hanley 2005, Chen et al. 2006), but not with large
FOV images (IRS, SPOT, Landsat, MERIS), due their inability to model high-frequency
distortions inherent to raw level-1A large-swath or long-strip images (Madani 1999,
Dowman and Dolloff 2000). On the other hand, 3D physical models achieving (sub-) pixel
accuracy were largely developed for all types of EO sensors (airborne and spaceborne, low
to high resolution) since the 1970s.
4.1.2 Geo-referenced ‘level 1B’ images

Since they have been systematically corrected and geo-referenced, the ‘level 1B’ images just
retain the terrain elevation distortion, in addition to a rotation-translation related to the
map reference system. A 3D 1st-degree polynomial model with Z-elevation parameters
could thus be efficient depending of the requested final accuracy. For scanners with across-
track viewing capability, only the Z-elevation parameter in the X-equation is useful. The
2nd-degree polynomial models could also be used (Palà and Pons 1995) for correcting
some residual errors of the 1B processing. When possible, solutions to overcome the
empirical model approximation are either to convert the 1B-images back into 1A-images
using the metadata and the reverse transformation (Al-Roussan et al. 1997), or to
‘re-shape and re-size’ the 1B-images to the raw imagery format (Valadan Zoej and Petrie
1998). This 1B-geometric modelling can be mathematically fused with normal level-1A
3D physical model to avoid multiple image resampling.
4.1.3 Map-oriented ‘level 2A’ images

The map-oriented images (‘level 2A’) also retain the elevation distortion but image lines
and columns are no more related to sensor-viewing and satellite directions. A 3D 1st-
degree polynomial model with Z-elevation parameter in both axes can thus be efficient
depending of the requested final accuracy. Such as for level 1B, 2nd-degree empirical
models (polynomial or rational) can be used for correcting some residual errors of the 2A
processing. Due to the fact that IKONOS Geo images were already corrected for all
systematic geometric distortions, except the terrain relief, 3D empirical models were
applied since 2000 to achieve pixel accuracy or better (Grodecki 2001).
The 3D physical model has been still approximated and developed for IKONOS Geo
images using basic information of the metadata and celestial mechanics laws (Toutin
2003a) and was largely adopted by the photogrammetric and RS community (Davis and
Wang 2001, Wang and Ellis 2005 and many others). Even approximated, this 3D physical
model (using the fusion of the geometric distortions and their mathematical functions) has
been proven to be robust and to achieve consistent results over different study sites, relief,
environments and cartographic data (Toutin 2003a, Wang and Ellis 2005).
4.1.4 SAR images

SAR images are standard products in slant or ground range presentations. They are
generated digitally during post-processing from the raw signal SAR data (Doppler
frequency, time delay). Errors present in the input parameters related to image geometry
model will propagate through to the image data. These include errors in the estimation of
slant range and of Doppler frequency and also errors related to the satellite’s ephemeris
and the ellipsoid. Assuming the presence of some geometric error residuals, the parameters
of a 3D physical model reflect these residuals. Furthermore, since different 3D SAR
physical models are largely available (Curlander 1982, Leberl 1990, Schreier 1993), few
attempts have been performed to apply 3D polynomial or RF empirical models to SAR
images (spaceborne or airborne). More recently with the launch of the new SAR systems,
MDA provided with Radarsat-2 image (high-to-low resolution, slant or ground range)
3rd-degree RFMs, which have accuracy similar to 3D MDA physical model used to
generate them (Robertson 2009, personal communications), but with larger errors with
ScanSAR and steep angles. These Radarsat-2 RFMs were independently evaluated and
validated with a radargrammetric model (Toutin and Omari 2011). Taking advantages of
the two models (physical and empirical) they fused them to generate a new hybrid model,
which can be applied to Radarsat-2 data without any user’s GCP. Zhang et al. (2010),
using a range/Doppler model, computed an adaptive RFM (terrain-independent RFM
approach) with various degrees to numerator and denominator polynomial functions and
applied it to different SAR sensors (TerraSAR, Radarsat-2, Cosmo-SkyMed, ALOS, etc.)
with a good accuracy compared to their range/Doppler model. Nevertheless, a determin-
istic model is still required to compute the empirical RFM.
4.2 Acquisition of GCPs

Whatever the VIR and/or SAR geometric model used, even RF ‘terrain-independent’
approach, some GCPs have to be acquired to compute/refine the parameters of
the mathematical functions in order to obtain sub-pixel accuracy. The number of GCPs
20 T. Toutin
is a function of different aspects: the method of collection, the sensor type and resolution,
the image spacing, the geometric model, the study site, the physical environment, GCP
definition and accuracy and the final expected accuracy. Figure 5(a) are examples of well-
defined GCPs and tools to extract image coordinates. For HR SAR images, metallic
structures (electrical poles, Figure 5b) are good candidates for GCPs.
Consequently, all the aspects of GCP collection have to be considered as a whole, and
not separately, to avoid too large discrepancies in accuracy between these different aspects.
For example, differential GPS survey with decimetre accuracy is too accurate (and
expensive in remote areas) to process SPOT data (3–10 m resolution), while road
intersections and 1 : 20–50,000 topographic maps are not accurate enough to process
QuickBird images if you expect sub-pixel (decimetre) accuracy, etc. The weakest aspect in
GCP collection, which is of course different for each study site and image, will thus be the
major source of error in the error budget and propagation in the math model.
Furthermore, to ensure robustness and consistency in an operational environment, it is
safer to collect more than the minimum required depending of the math model used and to
have medium-accurate GCPs than no GCP in remote areas.
With 3D 1st-approach RF models provided in metadata, 1–10 GCPs are needed to
remove the remaining errors with 2D polynomial functions (0–2nd degree) or to refine
the RF parameters. With 3D physical models, fewer GCPs (0–6) are generally required
per image. There is, however, no relative orientation between adjacent images because
RFMs were computed independently by the image providers. When GCP accuracy is
worse than the pixel size, its number should be increased, depending also of the final
expected accuracy (Savopol et al. 1994, Wang and Ellis 2005, Fraser et al. 2006). With
deterministic models, GCPs should be only acquired at the border of the image and
cover the full elevation range of the terrain to avoid extrapolation in planimetry and
elevation, without a regular distribution. On the other hand, since empirical models can
be sensitive to GCP distribution, number and accuracy, GCPs should be spread over
the full image(s) in planimetry and also in the elevation range to avoid large errors
between GCPs.
4.3 Geometric model computation

When multi-images path or block are processed over large study sites, a spatio-
triangulation based on a block adjustment is performed to compute all geometric models
simultaneously by a least-squares adjustment, so that the individual models are properly
tied and the entire block is optimally oriented in relation to the GCPs. The spatio-
triangulation method has been applied using 3D physical models to different VIR/SAR/
HR data: acquired from either single sensor (Leberl 1990, Veillet 1991, Belgued et al. 2000,
Campagne 2000, Kornus et al. 2000, Toutin 2003b–e, Bouillon et al. 2006) or multiple
sensors (Toutin 2004, 2006c); as well as using 3D RF models to single HR optical sensor
(Grodecki and Dial 2003).
With the spatio-triangulation, the same number of GCPs is theoretically needed to
adjust a single image, an image strip or a block: it drastically reduces the number of GCPs
(Bouillon et al. 2006). Tie points (TPs) between the adjacent images are thus used to
link the images and/or strips. Elevation of TPs (ETPs) must be added when the
intersection geometry of the adjacent images is weak, such as with intersection angles
Figure 5. (a and b) Examples of well-defined GCPs and tools for image pointing and extracting their
image coordinates for HR optical and SAR images, respectively.
22 T. Toutin
less than 15–20 (Toutin 2003c–e). There are a number of advantages to the
spatio-triangulation process, namely:
. the reduction of the number of GCPs;
. a better relative accuracy between the images;
. a more homogeneous and precise mosaic over large areas and
. a homogeneous GCP network for future geometric processing.
Whatever the number of images and the geometric models used, each GCP contributes
to two observation equations for each image: an equation in X and an equation in Y. The
observation equations are used to establish the error equations for GCPs, TPs and ETPs.
Each group of error equations can be weighted as a function of the accuracy of the image
and cartographic data. The normal equations are then derived and resolved with the
unknowns computed. In addition for the 3D physical models, conditions or constraints on
osculatory orbital parameters or other parameters (GPS/INS) can be added in the
adjustment to take into account the knowledge and the accuracy of the ephemeris or other
data, when available. They thus prevent the adjustment from diverging and they also filter
the input errors.
Since there are always redundant observations to reduce the input error propagation in
the geometric models a least-square adjustment is generally used. When the mathematical
equations are non-linear, which is the case for physical and 2nd and higher degree
empirical models, some means of linearisation (series expansions or Taylor’s series) must
be used. A set of approximate values for the unknown parameters in the equations must be
thus initialised:
. to zero for the empirical models, because they do not reflect the image acquisition
geometry; or
. from the osculatory orbital/flight and sensor parameters of each image for the
physical models.
More information on least-squares methods applied to geomatics data can be obtained
in Mikhail (1976) and Wong (1980). The results of this processing step are:
. the parameter values for the geometric model used for each image;
. the residuals in X and Y directions (and Z if more than one image is processed) for
each GCP/ETP/TP and their root mean square (RMS) residuals;
. the errors in X and Y directions (and Z if more than one image is processed) for
each ICP if any, and their bias and RMS errors and
. the computed cartographic coordinates for each point, including ETPs and TPs.
When more GCPs than the minimum theoretically required are used, the GCP
residuals reflect the modelling accuracy, while the ICP RMS errors reflect the final
accuracy taking into account ICP accuracy. When GCPs/ICPs are not accurate, their
errors are included in the computed RMS residuals/errors, respectively. By using of
overabundant GCPs with physical models, the input data errors (image measurement and/
or map) do not propagate through the physical models but are mainly reflected in the GCP
residuals due to a global adjustment. Consequently, it is thus ‘normal and safe’ with 3D
physical models to obtain RMS residuals in the same order of magnitude than the GCP
accuracy, but the physical model by itself will be more precise and the internal accuracy of
image(s) will be better than the RMS residuals.
4.4 Ortho-rectification
The last step of the geometric processing is the image rectification, preferably with DEM
(Figure 6). To ortho-rectify the uncorrected image into a map-referenced image, there are
two processing operations:
. a geometric operation to compute the cell coordinates in the uncorrected image
for each map image cell; and
. a radiometric operation to compute the intensity value or digital number (DN) of
the map image cell as a function of the intensity values of uncorrected image cells
that surround the previously computed position of the map image cell.
4.4.1 Geometric operation

The geometric operation requires the geometric math model and preferably elevation
information for a better geopositioning. For any map coordinates (X, Y) of the rectified
image in the map system and Z-elevation extracted from DEM if necessary, the coordinates
(column and line) in the uncorrected image is computed from the previously resolved math
model. Because these computed coordinates will not overlay in the pixel centre of the
uncorrected image, the column and line computed values are never integer values.
The final grid spacing of the rectified image will be generally close to the sensor
resolution. When data fusion from multi-sensor images with different resolutions are
performed in the map reference system, the common grid spacing requires geometric/
radiometric compromises. The features, which are geometrically and radiometrically
Figure 6. Image rectification to project the uncorrected image into the ground reference system:
the geometric (left) and radiometric (right) operations.
24 T. Toutin
Figure 7. Relationship between the DEM accuracy (in metres), the look angle (in degrees) of the
SAR image, and the resulting positioning error (in metres) generated on the SAR ortho-image.
The different boxes at the bottom represent the range of look angles for each Radarsat beam mode
(Toutin 1998).
distinguishable in the uncorrected images, should also be after the data fusion in the
ortho-rectified images. When the co-registration from the bundle adjustment can ensure
sub-pixel accuracy, the lowest grid spacing should be applied to keep the spatial integrity
of the highest resolution data. On the other hand, a geometric accuracy of one pixel and
more can generate in the data fusion colour artefacts and erroneous information
extraction due to mixed pixels. Consequently, to preserve the spatial integrity of the
highest resolution data degrades the spectral integrity of the fused images. Consequently,
the choice of the common grid spacing for the data fusion should be determined as
a function of the geometric accuracy to avoid mixed pixels, and of the information content
to be extracted.
Since the 2D models do not use elevation information, the accuracy of the resulting
rectified image (without DEM) will depend on the image viewing/look angle and the
terrain relief (Figures 7 and 8). On the other hand, 3D models take into account the
elevation, a DEM is thus fused with the uncorrected image through the mathematical
model to create accurate ortho-rectified images. This rectification should then be called an
ortho-rectification. But if no DEM is available, different altitude levels can be input for
different parts of the image (a kind of ‘rough’ DEM) to minimise the elevation distortion.
It is then important to have a quantitative evaluation of the DEM impact on the
rectification/ortho-rectification process, both in term of elevation accuracy for the
positioning accuracy and grid spacing for the level of details. This last aspect is more
important with HR images because a poor DEM grid spacing when compared to the
image spacing could generate artefacts for linear features (wiggly roads or edges).
Figures 7 and 8 give the relationship between the DEM accuracy (including
interpolation in the grid), the viewing and look angles with the resulting positioning
error on VIR and SAR ortho-images, respectively. These curves were mathematically
Figure 8. Examples of geometric resampling kernels applied to WorldView-1 panchromatic mode

image during the ortho-rectification process with DEM. The sub-images are 193 219 pixels with
0.15-m spacing. Letters A, B, C and D refer to different geometric resampling kernels (nearest
neighbour, bilinear, cubic convolution, sin(x)/x with 16 16 window), respectively. WorldView-1
Imageß and courtesy Digital Globe, 2009.
computed with the elevation distortion parameters of 3D physical model (Toutin 1995)
and can be simplified with the tangent of VIR viewing or SAR look angles, respectively.
The combined influence of the viewing angle, the relief height and the DEM characteristics
was also performed with HR satellite imagery (Zhang et al. 2001).
An other advantage of these curves is that they can be used to find any 3rd parameter
when the other two are known. For example (Figure 7), using SPOT image acquired with
10 viewing angle and 45-m accurate DEM, the error generated on the ortho-image is 9 m.
Inversely, if 4-m geo-positioning accuracy for the ortho-image is required and only 10-m
accurate DEM are available, the VIR image should be ordered with a viewing angle less
than 20 . The same error evaluation can be applied to SAR data using the similarly
computed curves (Figure 8). As other example, if positioning errors of 60 and 20 m on
standard-1 (S1) and fine-5 (F5) ortho-images, respectively, are required, a 20 m elevation
error, which includes the DEM accuracy and the interpolation into the DEM, is thus
sufficient. For HR images (spaceborne or airborne), the surface heights (buildings, forest,
26 T. Toutin
hedges) should be either included in the DTM to generate a digital surface model (DSM)
or taken into account in the overall elevation error. In addition, an inappropriate DEM in
term of grid spacing can generate artefacts with HR images acquired with large viewing
angles, principally over high relief areas (Zhang et al. 2001).
4.4.2 Radiometric operation

Since the computed coordinate values in the original image are not integer, one must
compute the DN to be assigned to the map image cell. In order to compute the DN to be
assigned to the map image cell, the radiometric operation uses a resampling kernel applied
to original image cells: either the DN of the closest cell (called nearest neighbour
resampling) or a specific interpolation or deconvolution algorithm using the DNs of
surrounding cells. In the first case, the radiometry of the original image and the image
spectral signatures are not altered, but the visual quality of the image is degraded.
In addition to the radiometric degradation, a geometric error of up to half pixel is also
introduced. This can caused a disjointed appearance in the map image. If these visual and
geometric degradations are acceptable for the end user, it can be an advantageous
solution.
In the second case, different interpolation or deconvolution algorithms (bilinear
interpolation or sinusoidal function) can be applied. The bilinear interpolation takes into
account the four cells surrounding the cell. The final DN is then either computed from two
successive linear interpolations in line and column using DNs of the two surrounding cells
in each direction or in one linear interpolation using DNs of the four surrounding cells.
The DNs are weighted as a function of the cell distance from the computed coordinate
values. Due to the weighting function this interpolation creates a smoothing in the final
map image.
The theoretically ideal deconvolution function is the sin(x)/x function. As this
sin(x)/x function has in infinite domain it cannot be exactly computed. Instead, it can be
represented by piecewise cubic function, such as the well-known cubic convolution.
The cubic convolution then computes 3rd-degree polynomial functions using a 4 4
cell window. DNs are first computed successively in the four-column and -line direction,
and the final DN is an arithmetical mean of these DNs. This cubic convolution
does not smooth, but enhances and generates some contrast in the map image
(Kalman 1985).
Due to computer improvement these last years, the sin(x)/x function can now be
directly applied as deconvolution function with different window sizes (generally 8 8 or
16 16). The computation time with 16 16 cell window can be 40–80 times more than
the computation time for nearest neighbour resampling. The finale image is, of course,
sharper with more details on features.
All these interpolation or deconvolution functions can be applied to VIR or SAR
images. However, they are geometric resampling kernels, not very well-adapted to SAR
images. Instead, it is better to use statistical functions based on the characteristics of the
radar, such as existing adaptive filters using local statistics (Lee 1980, Lopes et al. 1993,
Touzi 2002). Combining the filtering with the resampling also avoids multiple radiometric
processing and transformation, which largely degrades the image content and its
interpretation.
Since interpolation or deconvolution functions transform the DNs and then alter
the radiometry of the original image, problems may be encountered in subsequent
Figure 9. Examples of geometric/statistical resampling kernels applied to Radarsat-2 SAR ultra-fine

mode (U2) image during the ortho-rectification process with DEM. The sub-images are 265 262
pixels with 0.5-m spacing. Letters A, B, C and D refer to geometric resampling kernels (nearest
neighbour, bilinear, cubic convolution, sin(x)/x with 16 16 window), respectively, and letters E and
F refer to statistical adaptive SAR filters (Enhanced Lee and Gamma with 5 5 window),
respectively. ‘Radarsat-2 Dataß MacDonald, Dettwiler and Associates Ltd. (2008) – All Rights
Reserved’ and Courtesy of Canadian Space Agency.
spectral signature or pattern recognition analysis. Consequently, any process based on the
image radiometry should be performed before using interpolation or deconvolution
algorithms.
Figures 9 and 10 are examples of the application of different resampling kernels
applied to WorldView-1 panchromatic mode image and Radarsat-2 SAR ultra-fine mode
(U2) ground-range image (Toutin and Chénier 2009), respectively, during their ortho-
rectification process with DEM. Sub-images (around 200 200 pixels) were resampled
with a factor of three to better illustrate the variations of the resampling kernels: the
28 T. Toutin
Figure 10. Relationship between the DEM accuracy (in metres) the viewing angle (in degrees) of the
VIR image, and the resulting positioning error (in metres) generated on the ortho-image
(Toutin 1995).
WorldView and Radarsat resampled image pixels are then 0.15 and 0.5 m, respectively.
Letters A, B, C and D refer to different geometric resampling kernels (nearest neighbour,
bilinear, cubic convolution, sin(x)/x with 16 16 window), respectively, and letters E and
F refer to statistical adaptive SAR filters (Enhanced Lee and Gamma with 5 5 window),
respectively. For both VIR and SAR images, the nearest neighbour resampling kernel (A)
generates ‘blocky’ images, while the bilinear resampling kernel (B) generates fuzzy images
with the feeling of ‘out-of-focus’ images. One can also notice the step gradient, which are
reduced with the kernels sequence (from A to D). The best results are obtained with the
sinusoidal resampling kernels (C and D) but the true sinusoidal function (D) generates
sharper features well-representing the original circle shapes (Figure 9). The sin(x)/x with
16 16 window kernel should thus be favoured for optical images during the ortho-
rectification, even if the processing time is longer.
On the other hand with the Radarsat-2 SAR image (Figure 10), all the white dots
(corresponding to houses) change from square shapes (A) to round shapes (B, C, D),
which do not correspond to the original geometry of square houses. In addition, worm-
shape artefacts (typical from large oversampling) are generated in homogeneous flat areas
around the houses and in the ‘black’ streets; these artefacts become more evident with the
sequence of A–D resampling kernels. Consequently, these three last geometric resampling
kernels (bi-linear, cubic convolution and sin(x)/x) should not be used with SAR images.
The two adaptive filters (E and F) give not only a better image appearance than all
geometric resampling kernel due to the fact that the SAR speckle is filtered at the same
time, but keep the original feature shapes (even the neighboured houses are better
separated and discriminated) without generating worm-shape artefacts. These statistical
kernels are better adapted to SAR images, even with six-time oversampling of SAR
resolution (3 m), should always be favoured during the ortho-rectification.
5. Concluding remarks
Since the launch of the first civilian RS satellite in the 1970s, the needs and requirements
for geometrically process RS images have drastically changed due to the necessity of fusing
the different data. Furthermore, the fusion of multi-format data in a digital world requires
the highest accuracy possible so as to perform the final ortho-rectification of multi-sensor
images. Consequently, the full geometric correction process was evaluated with a data
fusion perspective: from the different input RS data from the imaging instruments (optical,
SAR, air- and space-borne, panchromatic, multi-band and -wavelength), their metadata
extracted from the non-imaging sensors (GPS, INS, star trackers and others), the
geographic and cartographic data (ground points, contour lines or DEM, maps, GIS, etc.),
the different geometric distortions and their associated math models to correct them until
the final ortho-rectification, which fused almost all the previous information into
a single product. The major advantage of this data fusion perspective is that the geometric
correction is considered as a whole procedure instead of looking, and addressing
separately and independently the different steps of this full and complex procedure. The
users can also better manage all data and information (acquisition, fusion, processing, etc.)
within the whole procedure (evaluation of the intermediate steps, error tracking, etc.) as a
function of the expected products and the desired result accuracy. This will result for the
users in a more appropriate and efficient exploitation of their multi-sensor, multi-source,
multi-format data for the specific thematic applications in development.
30 T. Toutin
Acknowledgement
Earth Sciences Sector (ESS) contribution number: 20100290.
References
Ahn, C.-H., Cho, S.-I., and Jeon, J.C., 2001. Orthorectification software applicable for IKONOS
high-resolution images: Geopixel-Ortho. Proceedings of IGARSS, 9–13 July Sydney,
Australia. Piscataway, NJ: IEEE, 555–557.
Al-Roussan, N., et al., 1997. Automated DEM extraction and ortho-image generation from SPOT
level-1B imagery. Photogrammetric Engineering and Remote Sensing, 63, 965–974.
Baetslé, P.L., 1966. Conformal transformations in three dimensions. Photogrammetric Engineering,
32, 816–824.
Bähr, H.P., 1976. Geometrische Modelle für Abtasteraufzeichnungen von Erkundungdsatlliten.
Bildmessung und Luftbildwesen, 44, 198–202.
Baltsavias, E.P. and Stallmann, D., 1992. Metric information extraction from SPOT images and the
role of polynomial mapping functions. International Archives of Photogrammetry and Remote
Sensing, 29 (B4), 358–364.
Bannari, A., et al., 1995. A theoretical review of different mathematical models of geometric
corrections applied to remote sensing images. Remote Sensing Reviews, 13, 27–47.
Belgued, Y., et al., 2000. Geometrical block adjustment of multi-sensor radar images. Proceedings of
EARSeL workshop: fusion of earth data, 26–28 January 2000 Sophia-Antipolis, France. Nice:
SEE GréCA/EARSeL, 11–16.
Bianconi, M., et al., 2008. A new strategy for rational polynomial coefficients generation. In:
C. Jürgens, ed. Proceedings of the EARSeL joint workshop, 5–7 March, Bochum, Germany,
21–28.
Bonneval, H., 1972. Levés topographiques par photogrammétrie aérienne, dans Photogramme´trie
ge´ne´rale: Tome 3, Collection scientifique de l’Institut Géographique National. Paris: Eyrolles
Editeur.
Bouillon, A., et al., 2002. SPOT5 HRG and HRS first in-flight geometric quality results. Proceedings
of SPIE, vol. 4881A sensors, system, and next generation satellites VII, 22–27 September Agia
Pelagia, Crete, Greece. Bellingham, WA: SPIE, CD-ROM (Paper 4881A-31).
Bouillon, A., et al., 2006. SPOT5 HRS geometric performance: using block adjustment as key issue
to improve quality of DEM generation. ISPRS Journal of Photogrammetry and Remote
Sensing, 60 (3), 134–146.
Campagne, Ph., 2000. Apport de la spatio-triangulation pour la caractérisation de la qualité image
géométrique. Bulletin de la Socie´te´ Française de Photogramme´trie et de Te´le´de´tection, 159,
66–26.
Centre National d’Études Spatiales (CNES), 1980. Le mouvement du ve´hicule spatial en orbite.
Toulouse, France: CNES, 1031.
Chen, L.-C. and Teo, T.-A., 2002. Rigorous generation of orthophotos from EROS-A high
resolution satellite images. International Archives of Photogrametry and Remote Sensing and
Spatial Information Sciences (8–12 July, Ottawa, Canada. Ottawa, ON: Natural Resources
Canada) 34 (B4), 620–625.
Chen, L.-C., Teo, T.-A., and Liu, C.-L., 2006. The geometrical comparison of RSM and RFM for
FORMOSAT-2 satellite images. Photogrammetric Engineering and Remote Sensing, 72 (5),
573–579.
Cheng, P. and Toutin, Th., 1998. Unlocking the potential for IRS-1C data. Earth Observation
Magazine, 6 (3), 24–26.
Curlander, J.C., 1982. Location of spaceborne SAR imagery. IEEE Transaction of Geoscience and
Remote Sensing, 22, 106–112.
Davis, C.H. and Wang, X., 2001. Planimetric Accuracy of IKONOS 1-m Panchromatic Image
Products. Proceedings of the ASPRS Annual Conference, 23–27 April, St Louis, Missouri,
USA. Bethesda, USA: ASPRS, CD-ROM (unpaginated).
Derenyi, E.E. and Konecny, G., 1966. Infrared scan geometry. Photogrammetric Engineering, 32,
773–778.
Dowmann, I. and Dolloff, J., 2000. An evaluation of rational function for photogrammetric
restitution. International Archives of Photogrammetry and Remote Sensing (16–23 July,
Amsterdam, The Netherlands. Amsterdam: GITC) 33 (B3), 254–266.
Ebner, H. and Muller, F., 1986. Processing of digital three line imagery using a generalized model for
combined point determination. International Archives of Photogrammetry and Remote Sensing
(19–22 August Rovamieni, Finland. Rovamieni: ISPRS) 26 (B3), 212–222.
Escobal, P.R., 1965. Methods of orbit determination. Malabar, USA: Krieger, 1–479.
Fraser, C.S., Baltsavias, E., and Gruen, A., 2002b. Processing of IKONOS imagery for sub-metre
3D positioning and building extraction. ISPRS Journal of Photogrammetry and Remote
Sensing, 56, 177–194.
Fraser, C.S., Dial, G., and Grodecki, J., 2006. Sensor orientation via RPCs. ISPRS Journal of
Photogrammetry and Remote Sensing, 60 (3), 182–194.
Fraser, C.S. and Hanley, H.B., 2005. Bias-compensated RPCs for sensor orientation of high-
resolution satellite imagery. Photogrammetric Engineering and Remote Sensing, 71 (8),
909–915.
Fraser, C.S., Hanley, H.B., and Yamakawa, T., 2002a. Three-dimensional geopositioning accuracy
of IKONOS imagery. Photogrammetric Record, 17, 465–479.
Fraser, C.S. and Ravanbakhsh, M., 2009. Georeferencing accuracy of GeoEye-1 imagery.
Photogrammetric Engineering and Remote Sensing, 75 (6), 634–638.
Friedmann, D.E., et al., 1983. Multiple scene precision rectification of spaceborne imagery with few
control points. Photogrammetric Engineering and Remote Sensing, 49, 1657–1667.
Gracie, G., et al., 1970. Stereo radar analysis, US Engineer Topographic Laboratory, Report no.
FTR-1339-1, Ft. Belvoir, VA, USA.
Grodecki, J., 2001. IKONOS stereo feature extraction – RPC approach. Proceedings of the ASPRS
annual conference, 23–27 April St Louis, MO. Bethesda, MD: ASPRS, CD-ROM
(unpaginated).
Grodecki, J. and Dial, G., 2003. Block adjustment of high-resolution satellite images
described by rational polynomials. Photogrammetric Engineering and Remote Sensing,
69 (1), 59–68.
Gugan, D.J., 1987. Practical aspects of topographic mapping from SPOT imagery. Photogrammetric
Record, 12, 349–355.
Guichard, H., 1983. Etude théorique de la précision dans l’exploitation cartographique d’un satellite
à défilement: application à SPOT. Bulletin de la Socie´te´ Française de Photogramme´trie et de
Te´le´de´tection, 90, 15–26.
Guindon, B. and Adair, M., 1992. Analytic formulation of spaceborne SAR image geocoding and
value-added products generation procedures using digital elevation data. Canadian Journal of
Hargreaves, D. and Robertson, B., 2001. Review of quickbird-1/2 and orbview-3/4 products from
MacDonald Dettwiler processing systems. Proceedings of the ASPRS annual conference, 23–27
April St Louis, MO. Bethesda, MD: ASPRS, CD-ROM (unpaginated).
Hoogeboom, P., Binnenkade, P., and Veugen, L.M.M., 1984. An algorithm for radiometric and
geometric correction of digital SLAR data. IEEE Transactions on Geoscience and Remote
Sensing, 22, 570–576.
Kalman, L.S., 1985. Comparison of cubic-convolution interpolation and least-squares
restoration for resampling Landsat-MSS imagery. Proceedings of the 51st annual ASP-
ASCM convention: ‘‘Theodolite to Satellite’’, 10–15 March Washington, DC. Falls Church,
VA: ASP, 546–556.
32 T. Toutin
Kersten, T., et al., 2000. Ikonos-2 Cartera Geo–Erste geometrische Genauigkeitsuntersuchungen in

der Schweiz mit hochaufgeloesten Satellitendaten. Vermessung, Photogrammetrie,
Kulturtechnik, 8, 490–497.
Khizhnichenko, V.I., 1982. Co-ordinates transformation when geometrically correcting earth space
scanner images, (in Russian). Earth Exploration From Space, 5, 96–103.
Kim, T. and Dowman, I., 2006. Comparison of two physical sensor models for satellite images:
position-rotation model and orbit-attitude model. The Photogrammetric Record, 21 (114),
110–123.
Konecny, G., 1970. Metric problems in remote sensing. ITC Publications Series A, 50, 152–177.
Konecny, G., 1976. Mathematische Modelle und Verfahren zur geometrischen Auswertung von
Zeilenabtaster-Aufnahmen. Bildmessung und Luftbildwesen, 44, 188–197.
Konecny, G., 1979. Methods and possibilities for digital differential rectification. Photogrammetric
Engineering and Remote Sensing, 45, 727–734.
Konecny, G., Kruck E., and Lohmann, P., 1986. Ein universeller Ansatz für die geometrischen
Auswertung von CCD-Zeilenabtasteraufnahmen. Bildmessung und Luftbildwesen, 54, 139–146.
Konecny, G., Lohmann, P., Engel, H., and Kruck, E., 1987. Evaluation of SPOT imagery on
analytical instruments. Photogrammetric Engineering and Remote Sensing, 53, 1223–1230.
Kornus, W., Lehner, M., and Schroeder, M., 2000. Geometric in-flight calibration by block
adjustment using MOMS-2P 3-line-imagery of three intersecting stereo-strips. Bulletin de la
Socie´te´ Française de Photogramme´trie et de Te´le´de´tection, 159, 42–54.
Kratky, W., 1987. Rigorous stereophotogrammetric treatment of SPOT images. Comptes- rendus du
Colloque International sur SPOT-1: utilisation des images, bilans, résultats, November Paris,
France. Toulouse: CEPADUES Editions, 1281–1288.
Kratky, W., 1989. On-line aspects of stereophotogrammetric processing of SPOT images.
Photogrammetric Engineering and Remote Sensing, 55, 311–316.
La Prade, G.L., 1963. An analytical and experimental study of stereo for radar. Photogrammetric
Engineering, 29, 294–300.
Leberl, F.W., 1972. On model formation with remote sensing imagery. Österreichiches Zeitschrift für
Vermessungswesen, 2, 43–61.
Leberl, F.W., 1978. Satellite radargrammetric, Deutsche Geodactische Kommission, Munich,
Germany, Serie C, 239, 1–156.
Leberl, F.W., 1990. Radargrammetric image processing. Norwood, MA: Artech House, 1–595.
Lee, J.S., 1980. Digital image enhancement and noise filtering by use of local statistics. IEEE
Transactions on Pattern Analysis and Machine Intelligence, 2, 165–168.
Lee, J.-B., et al., 2002. Improvement the positional accuracy of the 3D terrain data extracted from
IKONOS-2 satellite imagery. International Archives of Photogrammetry and Remote Sensing
(9–13 September Graz, Austria. Graz: Institute for Computer Graphics and Vision) 34 (B3),
B142–B145.
Light, D.L., et al., 1980. Satellite photogrammetry. In: C.C. Slama, ed. Manual of Photogrammetry.
4th ed., Chapter XVII. Falls Church, VA: ASP, 883–977.
Lopes, A., et al., 1993. Structure detection and statistical adaptive speckle filtering in SAR images.
International Journal of Remote Sensing, 14, 1735–1758.
Lu, Z., et al., 2010. Radar image and data fusion for natural hazards characterisation. International
Journal of Image and Data Fusion, 1 (3), 217–242.
Madani, M., 1999. Real-time sensor-independent positioning by rational functions. Proceedings of
ISPRS workshop on direct versus indirect methods of sensor orientation, 25–26 November
Barcelona, Spain. Barcelona, Spain: ISPRS, 64–75.
Masson d’Autume, G. de, 1979. Le traitement géométrique des images de télédétection. Bulletin de la
Mikhail, E.M., 1976. Observations and least squares. New York: Harper & Row, 1–497.
Naraghi, M., Stromberg, W., and Daily, M., 1983. Geometric rectification of radar imagery using
digital elevation models. Photogrammetric Engineering and Remote Sensing, 49, 195–159.
OGC, 1999. The open GISTM abstract specifications: the earth imagery case, Vol. 7, Available
from: http://www.opengis.org/techno/specs/htm/
Okamoto, A., 1981. Orientation and construction of models. part III: mathematical basis of the
orientation problem of one-dimensional central perspective photographs. Photogrammetric
Okamoto, A., 1988. Orientation theory of CCD line-scanner images. International Archives of
Photogrammetry and Remote Sensing (3–9 July Kyoto, Japan: ISPRS) 27 (B3), 609–617.
Okamoto, A., et al., 1998. An alternative approach to the triangulation of SPOT imagery.
International Archives of Photogrammetry and Remote Sensing (7–10 September Stuttgart,
Germany. Stuttgart: German Society for Photogrammetry and Remote Sensing) 32 (B4),
457–462.
Palà, V. and Pons, X., 1995. Incorporation of relief in polynomial-based geometric corrections.
Palubinskas, G., Reinartz, P., and Bamler, R., 2010. Image acquisition geometry analysis for the
fusion of optical and radar remote sensing data. International Journal of Image and Data
Fusion, 1 (3), 271–282.
Poli, D., 2007. A rigorous model for spaceborne linear array sensors. Photogrammetric Engineering
and Remote Sensing, 73 (2), 187–196.
Radhadevi, P.V., Ramachandran, R., and Mohan Murali, A., 1998. Precision rectification of IRS-
1C PAN data using an orbit attitude model. ISPRS Journal of Photogrammetry and Remote
Sensing, 53 (5), 262–271.
Robertson, B., 2003. Rigorous geometric modeling and correction of quickbird imagery. Proceedings
of the international geoscience and remote sensing IGARSS 2003, 21–25 July 2003 Toulouse,
France. Toulouse: CNES, CD-ROM.
Salamonowicz, P.H., 1986. Satellite orientation and position for geometric correction of scanner
imagery. Photogrammetric Engineering and Remote Sensing, 52, 491–499.
Savopol, F., et al., 1994. La correction géométrique d’images satellitaires pour la Base nationale de
données topographiques. Geomatica, e´te´, 48, 193–207.
Sawada, N., et al., 1981. An analytic correction method for satellite mss geometric distortion.
Schreier, G., ed., 1993. SAR geocoding and systems. Karlsruhe: Wichmann.
Schut, G.H., 1966. Conformal transformations and polynomials. Photogrammetric Engineering, 32,
826–829.
Shu, N., 1987. Restitution géométrique des images spatiales par la méthode de l’équation de
colinéarité. Bulletin de la Socie´te´ Française de Photogramme´trie et de Te´le´de´tection, 105, 27–40.
Suri, S., Schwind, P., Uhl, J., and Reinartz, P., 2010. Modifications in the SIFT operator for
effective SAR image matching. International Journal of Image and Data Fusion, 1, 243–256.
Sylvander, S., et al., 2000. Vegetation geometrical image quality. Bulletin de la Socie´te´ Française de
Photogramme´trie et de Te´le´de´tection, 159, 59–65.
Tao, V. and Hu, Y., 2001. A comprehensive study of the rational function model for
photogrammetric processing. Photogrammetric Engineering and Remote Sensing, 67,
1347–1357.
Tao, V. and Hu, Y., 2002. 3D reconstruction methods based on the rational function model.
Tannous, I. and Pikeroen, B., 1994. Parametric modelling of spaceborne SAR image geometry.
Application: SEASAT/SPOT image registration. Photogrammetric Engineering and Remote
Sensing, 60, 755–766.
Tong, X.H., Liu, S.J., and Weng, Q.H., 2010. Bias-corrected RPCs for high accuracy positioning
of QuickBird images. ISPRS Journal of Photogrammetry and Remote Sensing, 65 (2), 218–226.
Toutin, Th., 1983. Analyse mathématique des possibilités cartographiques du satellite SPOT,
Me´moire du diplôme d’Etudes Approfondies, Ecole Nationale des Sciences Géodésiques Saint-
Mandé, 1–74.
34 T. Toutin
Toutin, Th., 1995. Multi-source data integration with an integrated and unified geometric modelling.
EARSeL Advances in Remote Sensing, 4 (2), 118–129.
Toutin, Th., 1998. Evaluation de la précision géométrique des images de RADARSAT. Journal
canadien de te´le´de´tection, 24, 80–88.
Toutin, Th., 2003a. Error tracking in IKONOS geometric processing using a 3D parametric
modelling. Photogrammetric Engineering and Remote Sensing, 69, 43–51.
Toutin, Th., 2003b. Block bundle adjustment of Ikonos in-track images. International Journal of
Toutin, Th., 2003c. Block bundle adjustment of Landsat7 ETMþ images over mountainous areas.
Toutin, Th., 2003d. Path processing and block adjustment with RADARSAT-1 SAR images. IEEE
Transactions on Geoscience and Remote Sensing, 41, 2320–2328.
Toutin, Th., 2003e. Compensation par segment et bloc d’images panchromatiques et multibandes de
SPOT. Journal canadien de te´le´de´tection, 29 (1), 36–42.
Toutin, Th., 2004. Spatiotriangulation with multi-sensor VIR/SAR images. IEEE Transactions on
Geoscience and Remote Sensing, 42 (10), 2096–2103.
Toutin, Th., 2006a. Comparison of DSMs generated from stereo hr images using 3D
physical or empirical models. Photogrammetric Engineering and Remote Sensing, 72 (5),
597–604.
Toutin, Th., 2006b. Generation of DSMs from SPOT-5 in-track HRS and across-track HRG stereo
data using spatiotriangulation and autocalibration, ISPRS Journal of Photogrammetry and
Remote Sensing, 60 (3), 177–181.
Toutin, Th., 2006c. Spatiotriangulation with multisensor HR stereo-images. IEEE Transactions on
Geoscience and Remote Sensing, 44 (2), 456–462.
Toutin, Th. and Chénier, R., 2009. 3-D radargrammetric modeling of RADARSAT-2 ultrafine
mode: preliminary results of the geometric calibration. IEEE-GRSL, 6 (2), 282–286.
Toutin, Th. and Omari, K., 2011. A new hybrid model for geometric processing of Radarsat-2 data
without user GCP. Photogrammetric Engineering and Remote Sensing, 76, (accepted).
Toutin, Th., Zakharov, I., and Schmitt, C., 2010. Fusion of Radarsat-2 polarimetric images for
improved stereo-radargrammetric DEM. International Journal of Image and Data Fusion,
1 (1), 67–82.
Touzi, R., 2002. A review of speckle filtering in the context of estimation theory. IEEE Transactions
on Geoscience and Remote Sensing, 40, 2392–2404.
Valadan Zoej, M.J. and Petrie, G., 1998. Mathematical modelling and accuracy testing of SPOT
Level-1B stereo-pairs. Photogrammetric Record, 16, 67–82.
Vassilopoulou, S., et al., 2002. Orthophoto generation using IKONOS imagery and high-resolution
DEM: a case study on volcanic hazard monitoring of Nisyros Island (Greece). ISPRS Journal
of Photogrammetry and Remote Sensing, 57, 24–38.
Veillet, I., 1991. Triangulation spatiale de blocs d’images SPOT. The`se de Doctorat, Observatoire de
Paris, Paris, 1–101.
Wang, H. and Ellis, E.C., 2005. Spatial accuracy of orthorectified IKONOS imagery and historical
aerial photographs across five sites in China. International Journal of Remote Sensing, 26 (9),
1893–1911.
Wang, J., et al., 2005. Evaluation and improvement of geo-positioning accuracy of IKONOS stereo
imagery. Journal of Surveying Engineering, 131 (2), 35–42.
Westin, T., 1990. Precision rectification of SPOT imagery. Photogrammetric Engineering and Remote
Sensing, 56, 247–253.
Westin, T., 2000. Geometric modelling of imagery from the MSU-SK conical scanner. Bulletin de la
Wong, K.W., 1975. Geometric and cartographic accuracy of ERTS-1 imagery. Photogrammetric
Wong, K.W., 1980. Basic mathematics of photogrammetry. In: C. Slama, ed. Manual of
Photogrammetry, 4th ed., Chapter II. Falls Church, VA: ASP, 37–101.
Wong, F., Orth, R., and Friedmann, D., 1981. The use of digital terrain model in the rectification of
satellite-borne imagery. Proceedings of the 15th international symposium on remote sensing of
environment, 11–15 May Ann Arbor, MI: ERIM, 653–662.
Yang, X., 2001. Piece-wise linear rational function approximation in digital photogrammetry.
Proceedings of the ASPRS annual conference, 23–27 April St Louis, MO. Bethesda, MD:
ASPRS, CD-ROM (unpaginated).
Zhang, J., 2010. Multi-source remote sensing data fusion: status and trends. International Journal of
Image and Data Fusion, 1 (1), 5–24.
Zhang, G., et al., 2010. Evaluation of the RPC model for spaceborne sar imagery. Photogrammetric
Engineering and Remote Sensing, 77 (6), 727–733.
Zhang, Y., Tao, V., and Mercer, B., 2001. Assessment of the influences of satellite viewing angle,
earth curvature, relief height, geographical location and DEM characteristics on planimetric
displacement of high-resolution satellite imagery. Proceedings of the ASPRS annual
conference, 23–27 April, St Louis, MO. Bethesda MD: ASPRS, CD-ROM (unpaginated).

State-Of-The-Art of Geometric Correction of Remote Sensing Data: A Data Fusion Perspective

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

State-Of-The-Art of Geometric Correction of Remote Sensing Data: A Data Fusion Perspective

Uploaded by

Copyright:

Available Formats

International Journal of Image and Data Fusion

Vol. 2, No. 1, March 2011, 3–35

State-of-the-art of geometric correction of remote sensing

Canada Centre for Remote Sensing, Natural Resources Canada,

The three-dimensional (3D) geometric processing and ortho-rectification of

ISSN 1947–9832 print/ISSN 1947–9824 online

electro-optical (EO) sensors resulting in more and better information extraction;

2. Taxonomy of geometric distortions

Category Sub-Category Description of Error Sources

2.1 Distortions related to the platform

2.2 Distortions related to the sensor

2.3 Distortions related to the Earth

3. Geometric modelling of distortions

3.1 2D/3D empirical models

(2) 3D polynomial functions such as:

(3) 3D RFs, such as:

3.1.1 3D polynomial functions

3.1.2 3D rational functions

3.1.2.1 Terrain-independent RFM approach. The first approach, inappropriately called

3.1.2.2 Terrain-dependent RFM approach. The second approach, called terrain-

3.2 Physical models

Figure 4. Geometry of viewing of different types of scanners.

m21 ðX X0 Þ þ m22 ðY Y0 Þ þ m23 ðZ Z0 Þ

Nevertheless, to compute one ‘fused’ parameter is theoretically more precise than

4. Methods, processing and errors

4.1 Acquisition of images and metadata

4.1.1 Raw ‘level 1A’ images

4.1.2 Geo-referenced ‘level 1B’ images

4.1.3 Map-oriented ‘level 2A’ images

4.1.4 SAR images

4.2 Acquisition of GCPs

4.3 Geometric model computation

4.4.1 Geometric operation

Figure 8. Examples of geometric resampling kernels applied to WorldView-1 panchromatic mode

4.4.2 Radiometric operation

Figure 9. Examples of geometric/statistical resampling kernels applied to Radarsat-2 SAR ultra-fine

Kersten, T., et al., 2000. Ikonos-2 Cartera Geo–Erste geometrische Genauigkeitsuntersuchungen in

You might also like