High Resolution SAR Imaging and Signal Processing

Paulo A.C. Marques

Instituto Superior de Engenharia de Lisboa / Instituto de Telecomunicações
ISEL-ADEETC – R. Conselheiro Emidio Navarro, 1, Lisboa
PORTUGAL
pmarques@isel.pt

ABSTRACT
Synthetic aperture radar (SAR) is an active microwave system typically carried by an aircraft or a satellite
that is able to produce high resolution images of a target area. This paper presents the basic principles of
SAR imaging from a signal processing perspective, starting from an approximation for the received phase
signal, for the sake of simplicity. It then introduces the wavefront image reconstruction algorithm which
takes into consideration the exact phase history of the received signal. Other SAR imaging algorithms and
imaging modes are also addressed.

1.0 INTRODUCTION
Ground surveillance radar is a useful tool for remote sensing. Airborne and spaceborne radar sensors are able
to quickly cover large areas of the ground and to produce high quality radar maps, in all weather, day and
night, see [1], [2].

The so-called strip-map SAR acquisition geometry, illustrated in Figure 1, is the most widely used. This
acquisition geometry provides data along a terrain strip parallel to the flight direction (azimuth or cross-range
direction). The radar travels at constant velocity in the azimuth direction and transmits wide-band microwave
pulses at regular intervals. The corresponding echoes are recorded. Many pulses are transmitted during the so
called integration time, i.e., the time the platform takes to travel the footprint cross-range length.

y
Azimuth
RADAR or cross-range

Swath
x
Ground range
Footprint

Figure 1: Strip-Map SAR Geometry.

STO-EN-SET-235 2-1
High Resolution SAR Imaging and Signal Processing

High resolution in range is obtained using traditional pulse compression techniques and is independent of the
antenna size (see, e.g., [2] Ch. 7]). Typical look angle values ranges from 20° to 60°. Small look angles lead
to degradation of the ground range resolution, while large look angles produce strong shadow effects. High
resolution in the azimuth direction is obtained via synthesizing a long array by taking advantage of the
platform movement.

The paper is organized as follows.

Section 2 is dedicated to strip-map SAR imaging. It starts by presenting a comparatively simplified SAR
imaging algorithm, which considers the range and cross-range domains as separable. The basic principle
consists in implementing two-dimensional matched filtering. This algorithm, although an approximation,
permits the focusing of SAR images with quality good enough for many applications.

It then proceeds with the wavefront reconstruction algorithm, which exploits the structure of the SAR signals
in the Fourier domain. This algorithm tackles very accurately and efficiently the SAR inversion problem.
It permits very high resolution imaging even in spaceborne applications. For a comprehensive treatment,
see [3]. This section finishes with a briefly description of other popular SAR imaging algorithms.

Section 3 is dedicated to the SAR imaging modes. In this section, besides the strip-map SAR imaging, other
imaging modes supported by more recent SAR systems are addressed.

Section 4 presents the final remarks.

2.0 SAR IMAGE FORMATION

2.1 SAR Imaging Considering Separable Range and Cross-Range Domains

For the sake of simplicity we will now consider separately range and cross-range processing. For this
approach to hold, it is assumed that the distance from the RADAR platform to the target area is much larger
than the antenna footprint in cross-range dimension.

2.1.1 Range Domain: Pulse Compression

A radar signal should permit to determine the range of a target and to distinguish it from others in its
neighbourhood. Such a signal should have very short range extent and be powerful enough to maintain the
necessary signal to noise ratio on the receiver. In theory, a signal such as 𝛿𝛿(𝑡𝑡) would conform to such
specifications. However this kind of signal is not practical since it demands very short pulse duration with
power enough to conform to SNR requirements. In practice, to obtain high resolution in the range direction,
SAR systems typically transmit an alternative kind of signal, i.e. chirp signals which, basically,
are frequency modulated waveforms. The basic idea is to transmit a relatively long pulse that has a
bandwidth corresponding to a short duration pulse.

Let us consider the chirp waveform with duration Tp and bandwidth 𝛽𝛽 = 𝛼𝛼𝛼𝛼, modulated in frequency,
with constant amplitude:
2
𝑠𝑠(𝑡𝑡) = 𝑒𝑒 𝑗𝑗𝛼𝛼𝛼𝛼 (1)

As shown in [4] the emitted pulse can be compressed on the reception by correlating the received signal with
a copy of the emitted pulse. The obtained resolution is given by:
𝑐𝑐
Δ𝑥𝑥 = . (2)
2𝛼𝛼𝑇𝑇𝑝𝑝

2-2 STO-EN-SET-235
 High Resolution SAR Imaging and Signal Processing

Therefore, larger pulse duration achieves better range resolution.

Figure 2 illustrates qualitatively the pulse compression technique. The simulation considers five point-like
targets at coordinates 30m, 70m, 100m, 130m, and 160m. Fig. 2 on top-left, shows the real part of the
emitted chirp in baseband; on top-right, illustrates the five simulated targets and their relative reflectivity at
the corresponding ranges; bottom-left, shows the received signal before pulse compression. As expected the
several target echoes overlap and their relative reflectivities and positions are not resolvable. Fig. 2, bottom-
right, shows the resulting signal after correlating with the emitted chirp. The pulse compression operation
permits the detection of five targets and the estimation of their relative reflectivities.

Figure 2: Pulse Compression Illustration.

2.1.2 Cross-Range Domain

For the sake of simplicity we will now analyse the SAR processing on the cross-range (or azimuth) direction,
decoupled from the range axis.

Let us consider the scenario present in Figure 3 where the radar illuminates a single target while travelling
along the azimuth direction. The received signal phase difference between the emitted and the echoed signals
due to such a target is:

∆𝜃𝜃 = 2𝜋𝜋𝜋𝜋𝑡𝑡𝐷𝐷 , (3)

STO-EN-SET-235 2- 3
High Resolution SAR Imaging and Signal Processing

where:

2𝑅𝑅(𝑦𝑦) (4)
𝑡𝑡𝐷𝐷 = ,
𝑐𝑐

and:

𝑦𝑦 = 𝑣𝑣𝑅𝑅 𝑡𝑡, (5)

with:

𝑐𝑐 = 𝑓𝑓𝑓𝑓. (6)

y= vRt

RADAR Azimuth

R0 R(y)

y
Figure 3: Azimuth Direction SAR Geometry.

By considering 𝑅𝑅0 ≫ 𝑦𝑦 the received signal phase difference can be approximated to:

4𝜋𝜋𝑅𝑅0 2𝜋𝜋𝑦𝑦 2
Δ𝜃𝜃 ≈ + , (7)
𝜆𝜆 2𝑅𝑅0

where 𝑅𝑅0 denotes the closest approach distance between the radar and the target.

Therefore, the received signal can be written as:

𝑠𝑠(𝑦𝑦) ≈ 𝑒𝑒 𝑗𝑗4𝜋𝜋𝑅𝑅0 ⁄𝜆𝜆 𝑒𝑒 𝑗𝑗2𝜋𝜋𝑦𝑦

2 ⁄𝜆𝜆𝑅𝑅
0 , (8)

2-4 STO-EN-SET-235
 High Resolution SAR Imaging and Signal Processing

where the second term is easily recognized as a chirp. Notice that this chirp, although due to the SAR
geometry, can be also compressed using the correlation operation as happens with the emitted pulse in the
range direction.

The azimuth resolution can be shown to be given by [4]:

Δ𝑦𝑦 = 𝐷𝐷⁄2, (9)

where D is the antenna length.

Therefore, smaller antenna leads to better resolution, in contrast to what would be achieved without any
processing.

A very important aspect of SAR is the sampling in the cross-range direction. According to Shannon’s
sampling theorem, the signal should be larger than the bandwidth of the cross-range chirp. Therefore, the
pulse repetition frequency (PRF) should obey to:

2𝑣𝑣𝑅𝑅
𝑃𝑃𝑃𝑃𝑃𝑃 ≥ , (10)
𝐷𝐷
where 𝑣𝑣𝑅𝑅 is the SAR platform velocity in the azimuth direction.

When range and cross-range directions can be decoupled it is possible to apply a simple SAR processor for
focusing the raw-data as illustrated in Figure 4 for strip-map geometry.

Figure 4: Simplified SAR Processor for Strip-Map Geometry. The range compressed data
is obtained by a convolution with the range reference chirp; the azimuth compressed
data is obtained by a convolution with the azimuth reference function.

STO-EN-SET-235 2- 5
High Resolution SAR Imaging and Signal Processing

Let us recall that without any processing the cross range resolution would be:

𝜆𝜆
Δ𝑦𝑦 = 𝑅𝑅 . (11)
𝐷𝐷 0
That is, the resolution degrades with the range of the target. After processing, the obtained resolution is
independent of the target area range and is given by:

𝐷𝐷
Δ𝑦𝑦 = . (12)
2
As a numerical example of the resolution enhancement gain by using SAR let us look at the parameters of
the European Remote Sensing (ERS-1) satellite detailed on Table 1.

Table 1: ERS-1 Parameters.

Parameter Value
Wavelength 𝜆𝜆 = 5.6𝑐𝑐𝑐𝑐
Pulse bandwidth 𝐵𝐵 = 15𝑀𝑀𝑀𝑀𝑀𝑀
Antenna size (in azimuth) 𝐷𝐷 = 10𝑚𝑚
Look angle 𝜃𝜃 = 23°
Range 𝑅𝑅 = 850𝑘𝑘𝑘𝑘

By using eq. (11) one can calculate a cross-range resolution of 4.8km. However, by using the motion of the
antenna to synthesize a larger aperture, Eq. (12), the cross-range resolution becomes approximately 5m,
i.e. a thousand times better!

2.2 Wavefront Reconstruction Algorithm

The algorithm presented in previous subsection is able to obtain SAR images with quality good enough for
many applications. It uses a simple approach based on the validity of the quadratic approximation on the
phase of the received signal. Under this assumption both dimensions can be processed separately.

For high accuracy SAR processing, however, one needs to take into consideration the exact phase history of
the received signal. In this context we will now present the so-called wavefront reconstruction algorithm
which exploits the structure of the SAR signals in the Fourier domain. This algorithm tackles very accurately
and efficiently the SAR inversion problem even when the quadratic approximation is not valid.

This subsection follows very closely [3], [5] and [6].

Figure 5 shows a strip-map synthetic aperture radar scenario in the plane defined by the cross-range and the
slant-range axes. Let us consider a radar platform moving with velocity 𝑣𝑣𝑅𝑅 along a line known as the
synthetic aperture domain or the cross-range domain. For a fixed platform position, 𝑦𝑦 = 𝑣𝑣𝑅𝑅 𝑡𝑡 ≡ 𝑢𝑢, the radar
transmits a wide-band pulse 𝑝𝑝(𝑡𝑡), where t is the fast-time domain, and echoed signal from the illuminated
area is recorded.

2-6 STO-EN-SET-235
 High Resolution SAR Imaging and Signal Processing

Figure 5: Considered SAR Scenario for 2-D Imaging with the Wavefront Algorithm.

Let us consider a target region composed of N static point-like targets with complex reflectivity 𝑓𝑓𝑛𝑛 , where
𝑛𝑛 = 0, … , 𝑁𝑁 − 1, and coordinates (𝑥𝑥𝑛𝑛 , 𝑦𝑦𝑛𝑛 ) as illustrated in Fig. 5.

The target function corresponding to this scenario is:

𝑁𝑁−1

𝑓𝑓(𝑥𝑥, 𝑦𝑦) ≡ � 𝑓𝑓𝑛𝑛 𝛿𝛿(𝑥𝑥 − 𝑥𝑥𝑛𝑛 , 𝑦𝑦 − 𝑦𝑦𝑛𝑛 ). (13)

𝑛𝑛=0

The corresponding 2-D Fourier transform is:

𝑁𝑁−1

𝐹𝐹�𝑘𝑘𝑥𝑥 , 𝑘𝑘𝑦𝑦 � = � 𝑓𝑓𝑛𝑛 𝑒𝑒 −𝑗𝑗𝑘𝑘𝑥𝑥 𝑥𝑥𝑛𝑛 𝑒𝑒 −𝑗𝑗𝑘𝑘𝑦𝑦 𝑦𝑦𝑛𝑛 , (14)

𝑛𝑛=0

where 𝑘𝑘𝑥𝑥 and 𝑘𝑘𝑦𝑦 denote the spatial frequency with respect to coordinate x and y, respectively.

Radar imaging aims at determining the target function. To achieve this goal, we start by looking at the signal
received by the radar.

The echoed signal when the platform is at position y=u, in the fast-time frequency domain and in the slow-
time spatial domain:
𝑁𝑁−1

𝑠𝑠(𝜔𝜔, 𝑢𝑢) = 𝑃𝑃(𝜔𝜔) � 𝑎𝑎(𝜔𝜔, 𝑦𝑦𝑛𝑛 − 𝑢𝑢)𝑓𝑓𝑛𝑛 𝑒𝑒 −𝑗𝑗2𝑘𝑘𝑟𝑟𝑛𝑛 , (15)

𝑛𝑛=0

STO-EN-SET-235 2- 7
High Resolution SAR Imaging and Signal Processing

where 𝑃𝑃(𝜔𝜔) is the low-pass equivalent Fourier transform of the emitted pulse 𝑝𝑝(𝑡𝑡), 𝑎𝑎(𝜔𝜔, 𝑢𝑢) is the low-pass
equivalent of the two-way antenna radiation pattern in the (𝜔𝜔, 𝑢𝑢) domain, assumed to be independent of the
slant-range coordinate x, 𝑘𝑘 ≡ (𝜔𝜔 + 𝜔𝜔0 )/𝑐𝑐 is the wavenumber, 𝑓𝑓0 = 𝜔𝜔0 /(2𝜋𝜋) is the carrier frequency, and c
is the speed of light. Symbol 𝑓𝑓𝑛𝑛 denotes the complex reflectivity of the nth target, assumed to be independent
of the aspect angle. Wave divergence is absorbed by 𝑓𝑓𝑛𝑛 . Symbol 𝑟𝑟𝑛𝑛 denotes the distance between the radar
and the nth target and is given by:

𝑟𝑟𝑛𝑛 = �𝑥𝑥𝑛𝑛2 + (𝑦𝑦𝑛𝑛 − 𝑢𝑢)2 . (16)

In [3] it is shown that the Fourier transform of 𝑠𝑠(𝜔𝜔, 𝑢𝑢), with respect to the slow-time domain, is:
𝑁𝑁−1
2 𝑥𝑥
−𝑗𝑗�4𝑘𝑘 2 −𝑘𝑘𝑢𝑢
𝑆𝑆(𝜔𝜔, 𝑘𝑘𝑢𝑢 ) = 𝐴𝐴(𝜔𝜔, 𝑘𝑘𝑢𝑢 )𝑃𝑃(𝜔𝜔) � 𝑓𝑓𝑛𝑛 𝑒𝑒 𝑛𝑛 −𝑗𝑗𝑘𝑘𝑢𝑢 𝑦𝑦𝑛𝑛
𝑒𝑒 , (17)
𝑛𝑛=0

where 𝐴𝐴(𝜔𝜔, 𝑘𝑘𝑢𝑢 ) is the two-way antenna radiation pattern in the 2D frequency domain.

Comparing (14) with (17), we can write:

𝑆𝑆(𝜔𝜔, 𝑘𝑘𝑢𝑢 )𝑃𝑃(𝜔𝜔)𝐹𝐹�𝑘𝑘𝑥𝑥 , 𝑘𝑘𝑦𝑦 �, (18)

where:

𝑘𝑘𝑥𝑥 ≡ �4𝑘𝑘 2 − 𝑘𝑘𝑢𝑢2 , (19)

and:

𝑘𝑘𝑦𝑦 ≡ 𝑘𝑘𝑢𝑢 . (20)

To estimate the target function, we adopt a matched filtering approach, which is the most frequently used in
SAR applications, as it is light from the computational point of view and robust to model mismatches.

The output of the matched filter is thus:

𝐹𝐹� �𝑘𝑘𝑥𝑥 , 𝑘𝑘𝑦𝑦 � = 𝐴𝐴∗ �𝑘𝑘𝑥𝑥 , 𝑘𝑘𝑦𝑦 �𝑃𝑃∗ �𝑘𝑘𝑥𝑥 , 𝑘𝑘𝑦𝑦 �𝑆𝑆�𝑘𝑘𝑥𝑥 , 𝑘𝑘𝑦𝑦 �, (21)

resulting:
𝑁𝑁−1
2 2
𝐹𝐹� �𝑘𝑘𝑥𝑥 , 𝑘𝑘𝑦𝑦 � = �𝐴𝐴�𝑘𝑘𝑥𝑥 , 𝑘𝑘𝑦𝑦 �� �𝑃𝑃�𝑘𝑘𝑥𝑥 , 𝑘𝑘𝑦𝑦 �� � 𝑓𝑓𝑛𝑛 𝑒𝑒 −𝑗𝑗𝑘𝑘𝑥𝑥 𝑥𝑥𝑛𝑛 𝑒𝑒 −𝑗𝑗𝑘𝑘𝑦𝑦 𝑦𝑦𝑛𝑛 . (22)
𝑛𝑛=0

In the spatial domain, we have:

𝑁𝑁−1

𝑓𝑓̂(𝑥𝑥, 𝑦𝑦) = � 𝑓𝑓𝑛𝑛 ℎ(𝑥𝑥 − 𝑥𝑥𝑛𝑛 , 𝑦𝑦 − 𝑦𝑦𝑛𝑛 ), (23)

𝑛𝑛=0

2-8 STO-EN-SET-235
 High Resolution SAR Imaging and Signal Processing

where ℎ(𝑥𝑥, 𝑦𝑦) denotes the SAR system transfer function:

−1 2 2
ℎ(𝑥𝑥, 𝑢𝑢) = 𝐹𝐹𝐹𝐹(𝑥𝑥,𝑢𝑢) ��𝐴𝐴�𝑘𝑘𝑥𝑥 , 𝑘𝑘𝑦𝑦 �� �𝑃𝑃�𝑘𝑘𝑥𝑥 , 𝑘𝑘𝑦𝑦 �� �. (24)

The wavefront reconstruction algorithm block diagram is presented in Figure 6 and is summarized by the
following step sequence:
1) Compute 𝑆𝑆(𝜔𝜔, 𝑘𝑘𝑢𝑢 ), the two-dimensional Fourier transform of the received signal 𝑠𝑠(𝑡𝑡, 𝑢𝑢).
2) Implement the matched filtering in the frequency domain, by multiplying 𝑆𝑆(𝜔𝜔, 𝑘𝑘𝑢𝑢 ) with
𝑆𝑆0∗ (𝜔𝜔, 𝑘𝑘𝑢𝑢 ) = 𝐴𝐴∗ (𝜔𝜔, 𝑘𝑘𝑢𝑢 )𝑃𝑃∗ (𝜔𝜔).
3) Implement the change of variables (19) and (20), the so-called Stolt interpolation [12].
4) Calculate the two-dimensional inverse Fourier transform (22), obtaining 𝑓𝑓̂(𝑥𝑥, 𝑦𝑦), the estimate of the
target function.

s (t , u )

Fourier Transform
(t , u ) → (ω , ku )

S (ω , ku )

Matched Filtering X S0* (ω , ku )

Stolt Interpolation
k y = ku k x = 4k − ku2

Inverse
Fourier Transform
( k x , k y ) → ( x, y )

fˆ ( x, y )

Figure 6: Wavefront Reconstruction Imaging Algorithm.

STO-EN-SET-235 2- 9
High Resolution SAR Imaging and Signal Processing

This imaging algorithm belongs to a class of algorithms often referred to as wavenumber domain or 𝜔𝜔 − 𝑘𝑘
processors. This kind of processors is not new and has been used to perform seismic signal processing for
imaging the substrata of Earth [7]. These ideas were later applied to imaging of SAR data [11].

2.3 Other SAR Imaging Algorithms

Most SAR imaging algorithms usually include three major steps:
• Range compression;
• Range cell migration compensation;
• Azimuth compression.

In the wavefront reconstruction algorithm just described, the range compression and (partial) azimuth
compression are done in step 2. The range cell migration compensation is dealt with in step 3, where the
Stolt interpolation decouples the spatial domains x and y.

A large number of SAR focusing algorithms, besides the wavefront reconstruction algorithm, have been
presented in the literature. To reduce the computational requirements, at expense of image quality, some of
the algorithms implement approximated range migration compensation or they do not implement it at all.

Herein we just present a brief summary of the three most popular focusing algorithms, besides the wavefront
reconstruction already presented, for strip-map SAR.

2.3.1 Range-Doppler Algorithm

This algorithm is probably the most well-known SAR focusing scheme [1]. It starts with the 1-D transform
of 𝑠𝑠(𝑡𝑡, 𝑢𝑢) to 𝑠𝑠(𝑡𝑡, 𝑘𝑘𝑢𝑢 ). The coupling between the range (time) and 𝑘𝑘𝑢𝑢 is removed by a coordinate remapping
via interpolation. The azimuth compression is then applied in the 𝑘𝑘_𝑢𝑢 domain. A single compression in the
range-Doppler domain for each range is then applied to all pixels at the same range. In systems where range
cell migration is small, the Fresnel approximation can be used without degrading too much the image quality
and the interpolation step can be ignored (see, e.g., [8]).

2.3.2 Chirp-Scaling Algorithm

The chirp-scaling algorithm starts with the raw-echo data (without range pulse compression) because it
exploits the linear FM (LFM) signal structure of the transmitted pulse [9]. If coded or pseudorandom
sequences are used for the transmitted pulse, data must be range compressed with an adequate matched filter
and then respread by a convolution with a LFM pulse. As in the range-Doppler approach, the algorithm starts
with a 1-D transform in the slow-time domain of 𝑠𝑠(𝑡𝑡, 𝑢𝑢)to 𝑠𝑠(𝑡𝑡, 𝑘𝑘𝑢𝑢 ). Then 𝑠𝑠(𝑡𝑡, 𝑘𝑘𝑢𝑢 ) is multiplied by a phase-
only chirp which scales the range chirp for each 𝑘𝑘𝑢𝑢 and 𝑡𝑡 so that the Doppler loci of all targets end up with a
common range migration in phase. The scaled data is then Fourier-transformed to (𝜔𝜔, 𝑘𝑘𝑢𝑢 ) domain, where
range compression and range cell migration compensation are performed. Azimuth compression and residual
phase compensation are then implemented in the (𝑡𝑡, 𝑘𝑘𝑢𝑢 ) domain.

2.3.3 Time Domain Correlation Algorithm

The basic principle behind the time domain correlation (TDC) imaging algorithm is simply the
implementation of the matched filtering via a two-dimensional correlation ([3], Ch. 4.7), similarly to
algorithm presented in subsection 2.1. To reduce the numerical errors the signal is up-sampled (interpolated).

The major drawback of this algorithm is the high computational cost due to the brute force approach.

2 - 10 STO-EN-SET-235
 High Resolution SAR Imaging and Signal Processing

3.0 SAR IMAGING MODES

Besides the more fundamental strip-map imaging mode which has been previously presented, there are many
others supported by the recent SAR systems. In order to support these imaging modes the antenna is divided
into several sub-apertures. By controlling the phase and amplitude of each subaperture it is possible to
change the antenna radiation pattern and, consequently, the illuminated region.

3.1 Strip-Map SAR

The strip-map SAR, which has been addressed in the previous sections, is the most common configuration.
In this imaging mode the antenna patterns remains fixed and it illuminates a fixed swath. The data
acquisition occurs for a continuous strip of the terrain as presented in Figure 7.

Flight
path

Swath

Figure 7: Strip-Map SAR Imaging Mode.

3.2 ScanSAR
The ScanSAR imaging mode is used when a wider swath is required. The antenna pattern is steered in
elevation during the data acquisition interval. This leads to the illumination of several sub-swaths as
illustrated in Figure 8. Since each sub-swath is illuminated during a fraction of time when compared with the
strip-map situation, the azimuth resolution is proportionally degraded.

STO-EN-SET-235 2- 11
High Resolution SAR Imaging and Signal Processing

Flight direction

Subswath 1

Subswath 2

Figure 8: ScanSAR Mode Increases the Swath Width at Expenses of Azimuth Resolution.

3.3 Spotlight SAR

The spotlight SAR imaging configuration is used to obtain the best resolution of a certain target area.
The resolution is enhanced by increasing the integration time. Figure 9 illustrates this imaging mode.
The antenna pattern is steered in order to illuminate a fixed region of interest during a much longer interval
than the one achieved in strip-map SAR operation. The increase in resolution is obtained at the cost of not
imaging a continuous swath along the flight path.

2 - 12 STO-EN-SET-235
 High Resolution SAR Imaging and Signal Processing

Flight
path

Swath

Figure 9: Spotlight Mode: Increases the Azimuth Resolution

Having as Trade-Off the Decrease of the Swath Length.

3.4 High-Resolution Wide-Swath Imaging

To overcome the trade-off limitation between resolution in azimuth versus spatial coverage, a novel imaging
technique has been proposed [10]. This imaging mode combines digital beamforming on receive in elevation
with multiple aperture recording in azimuth. This strategy enables the imaging of wide swath while
simultaneously obtaining high resolution.

Multiple Rx-channels

Wide swath
illumination

Figure 10: High-Resolution Wide-Swath SAR Simultaneously Improves Azimuth Resolution

and Swath Width at Expenses of Hardware and Computational Complexity.

STO-EN-SET-235 2- 13
High Resolution SAR Imaging and Signal Processing

This imaging mode is only made possible due to the more recent technological advances that allow for
digital beamforming techniques implementation with simultaneous multi-aperture signal acquisition. is
currently under development and implementation at EADS Astrium, being supported by the German
Aerospace Center (DLR) [13].

4.0 FINAL REMARKS

This paper is presents and discusses some of the most popular SAR geometries and imaging algorithms.
In particular, the strip-map geometry – one of the most widely used SAR data acquisition techniques –
was analysed. Firstly, a simple technique for SAR imaging, based on the possibility of processing range and
cross-range dimensions separately, was presented. Although it is an approximation it permits focusing SAR
images with sufficient quality for less demanding applications.

The framework of the wavefront reconstruction algorithm was then introduced. This algorithm is able to deal
with the coupling between fast-time and slow-time domains tackling very accurately and efficiently the SAR
inversion problem. It permits very high resolution imaging even in spaceborne applications.

Besides strip-map SAR, other popular SAR imaging modes were addressed. There is a trade-off limitation
between resolution in azimuth versus spatial coverage when considering strip-map SAR, ScanSAR,
or Spotlight SAR. To overcome this limitation the more advanced high-resolution wide-swath imaging SAR,
was briefly described. It combines digital beamforming with multiple aperture recording in azimuth,
enabling imaging of wide swath while simultaneously achieving high resolution.

5.0 ACKNOWLEDGEMENTS
The author wishes to acknowledge Prof. José-Bioucas Dias for the joint research activity carried together
along several years that resulted in some of the publications and techniques that are herein summarized.

6.0 REFERENCES
[1] M. Soumekh, Fourier Array Imaging. Prentice Hall, 1994.

[2] A. Rihaczec, Principles of High-Resolution Radar. Artech House, 1996.

[3] M. Soumekh, Synthetic Aperture Radar Signal Processing with MATLAB algorithms. WILEY-
INTERSCIENCE, 1999.

[4] D. C. Munson; R. L. Visentin, A signal processing view of strip-mapping synthetic aperture radar,
IEEE transactions on Acoustics, Speech, and Signal Processing, Vol. 37, Issue 12, pp. 2131-2147,
1987.

[5] P. Marques, Moving objects imaging and trajectory estimation using a single synthetic aperture radar
sensor, Ph.D. dissertation, Technical University of Lisbon, 2004.

[6] J. Bioucas-Dias, P. Marques, Multiple moving target detection and trajectory estimation using a single
SAR sensor, IEEE Transactions on Aerospace and Electronic Systems, Vol. 39, No. 2, pp. 604-624,
April 2003.

[7] J. Gazdag and P. Sguazzero, Migration of seismic data, Proceedings of the IEEE, Vol. 1, pp. 23-48,
1984.

2 - 14 STO-EN-SET-235
 High Resolution SAR Imaging and Signal Processing

[8] P. Gough and D. Hawkins, Imaging algorithms for a stripmap synthetic aperture sonar: Minimizing the
effects of aperture errors and aperture undersampling, IEEE Journal of Oceanic Engineering, Vol. 22,
No. 1, pp. 27-39, 1997.

[9] R. Raney, H. Runge, R. Bamler, I. Cumming, and F. Wong, Precision SAR processing using chirp
scaling, IEEE Transactions on Geoscience and Remote Sensing, Vol. 32, No. 4, pp. 786-799, 1994.

[10] G. D. Callaghan and I. Longstaff, “Wide-swath space-borne SAR using a quad-element array,”
IEE Proc. Radar, Sonar Navigation, Vol. 146, No. 3, pp. 155-165, 1999.

[11] C. Cafforio, C. Prati, and F. Rocca, “SAR data focusing using sysmic migration techniques,” IEEE
Transactions on Aerospace and Electronic Systems, Vol. 27, No. 2, pp. 194-206, March 1991.

[12] H. Stolt, “Migration by Fourier transform,” Geophysics, Vol. 1, pp. 23-48, February 1978.

[13] A. Moreira, “A golden age for spaceborne SAR systems”, International Conference on Microwaves,
Radar, and Wireless Communication, MIKON, 2014.

STO-EN-SET-235 2- 15
High Resolution SAR Imaging and Signal Processing

2 - 16 STO-EN-SET-235