burst-mode and scansar interferometry - geoscience and …€¦ · possible to map 80% of the...

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 40, NO. 9, SEPTEMBER 2002 1917

Burst-Mode and ScanSAR InterferometryJürgen Holzner, Student Member, IEEE,and Richard Bamler, Senior Member, IEEE

Abstract—ScanSAR interferometry is an attractive option forefficient topographic mapping of large areas and for monitoringof large-scale motions. Only ScanSAR interferometry made it pos-sible to map almost the entire landmass of the earth in the 11-dayShuttle Radar Topography Mission. Also the operational satellitesRADARSAT and ENVISAT offer ScanSAR imaging modes andthus allow for repeat-pass ScanSAR interferometry. This papergives a complete description of ScanSAR and burst-mode inter-ferometric signal properties and compares different processingalgorithms. The problems addressed are azimuth scanning patternsynchronization, spectral shift filtering in the presence of highsquint, Doppler centroid estimation, different phase-preservingScanSAR processing algorithms, ScanSAR interferogram forma-tion, coregistration, and beam alignment. Interferograms anddigital elevation models from RADARSAT ScanSAR Narrowmodes are presented. The novel “pack-and-go” algorithm forefficient burst-mode range processing and a new time-variant fastinterpolator for interferometric coregistration are introduced.

Index Terms—Burst-mode, ENVISAT, pack-and-go algorithm,RADARSAT, ScanSAR, SRTM, synthetic aperture radar inter-ferometry.

I. INTRODUCTION

SCANSAR uses the burst-mode technique to image largeswaths at the expense of azimuth resolution [1], [2]. Typical

swath widths of RADARSAT and ENVISAT/ASAR ScanSARimages are 300–500 km, and azimuth resolution is in the orderof 50 m–1 km. ENVISAT/ASAR uses burst-mode imaging alsofor its alternating polarization mode [3].

Interferometric (InSAR) applications can benefit from theincreased swath width of ScanSAR data or the increased in-formation content in alternating polarization mode data. Someattractive properties of wide swaths are more economic use ofground control points for calibration of the imaging geometry,reduced need for postprocessing to mosaic large digital eleva-tion models (DEMs), and more frequent coverage for differen-tial InSAR applications. Only ScanSAR interferometry made itpossible to map 80% of the earth’s landmass during the 11-dayShuttle Radar Topography Mission (SRTM).

In all of these cases, and in contrast to the traditionalstrip-map modes, the raw data consist of bursts of radar echoeswith azimuth extent shorter than the synthetic aperture length.This acquisition strategy is the reason for the azimuth-variant

Manuscript received December 18, 2001; revised June 20, 2002. This workwas carried out in the context of RADARSAT NASA-ADRO Project 500. TheRADARSAT data are copyright Canadian Space Agency 1996, 1997, and 1998.DLR’s research and development activities on RADARSAT ScanSAR interfer-ometry were supported in part by the former DARA in the framework of theGEMSAR project and by the German Canadian Scientific Technical Coopera-tion Programme.

The authors are with the German Aerospace Center (DLR), Oberpfaffen-hofen, D-82234 Wessling, Germany (e-mail: [email protected]).

Digital Object Identifier 10.1109/TGRS.2002.803848

spectral properties of the signal (Section III) and complicatessynthetic aperture radar (SAR) signal processing and interfer-ometric processing.

Since the ScanSAR modes of current satellites were moti-vated by radiometric imaging rather than by interferometry,the development of phase-preserving and interferometricburst-mode processors is far behind the established processingchains for strip-map mode data. There is, for example, nocomplex ScanSAR product in the official RADARSAT orENVISAT/ASAR product lists. Several algorithms have beenproposed in the scientific literature [4]–[7], and an operationaldedicated system for the single-pass SRTM interferometer hasbeen built up by the Jet Propulsion Laboratory recently [8]. Thepurpose of this paper is to give a description of burst-mode andScanSAR interferometry, including signal properties and nec-essary processing steps. Several algorithms are compared withrespect to their expected accuracy, throughput, implementationeffort, and reusability of software from strip-mode processors.The repeat-pass case is considered throughout the paper,and the theoretical findings are illustrated by a RADARSATScanSAR example.

The paper is organized as follows. Section II describesthe RADARSAT modes and the dataset used in the paper.Section III reviews the particular properties of burst-modeSAR data. Section IV discusses important preprocessing steps.In Section V several phase-preserving burst-mode processingalgorithms are compared, while Section VI discusses twointerferometric processing schemes. Section VII finally givesan example for a first RADARSAT ScanSAR DEM.

II. I NTERFEROMETRICRADARSAT DATASETS

The RADARSAT SAR instrument provides various imagingmodes, such as the strip-map mode with standard (S1–S7), wide(W1–W3), and fine (F1–F5) beams as well as various ScanSARmodes [9], [10]. The so-called ScanSAR Narrow modes com-bine beams W1 and W2 for the “ScanSAR Narrow Near,” orW2, S5, and S6 for the “ScanSAR Narrow Far” modes, respec-tively. Both narrow modes cover a swath width of about 300 km.The ScanSAR wide modes combine beams W1, W2, W3, andS7, or W1, W2, S5, and S6, providing swath widths of 500and 450 km, respectively. RADARSAT is the first spacebornesystem to operate in ScanSAR mode that is also able to providesuitable data for ScanSAR interferometry.

RADARSAT was not designed originally for repeat-passInSAR. Both the orbit accuracy and the repeat cycle of 24days are not favorable in this respect. On the other hand,RADARSAT’s 30-MHz fine-resolution mode combined withlarge incidence angles ( 49 ) provides an interferometricmode with a critical baseline as large as 6 km. First re-sults on interferometry and differential interferometry using

0196-2892/02$17.00 © 2002 IEEE

1918 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 40, NO. 9, SEPTEMBER 2002

Fig. 1. Critical baseline of various RADARSAT modes and ERS as a functionof incidence angle. The bandwidths of the SAR modes are 11.6, 17.3, 30.0, and15.5 MHz for wide, standard, fine beam mode, and ERS, respectively.

RADARSAT’s strip-map modes have been reported in [11].In ScanSAR modes, however, the range bandwidth is as lowas 11.6 MHz, giving a critical baseline in the order of 600 m(Fig. 1).

RADARSAT cannot synchronize its azimuth burst cycles atdifferent acquisitions. In order to maximize coherence, a stepcalled azimuth scanning pattern synchronization (ASPS) is re-quired for preprocessing the datasets. The aim is to preserveonly those raw data lines of companion bursts that image thesame area on ground (Section IV). Since the chance of ob-taining overlapping raw data lines decreases with the number ofbeams, only ScanSAR Narrow mode can be considered suitable.We have ordered data acquisitions based on RADARSAT’s pre-dicted orbit drifts. Among the 12 ScanSAR datasets, two pairscould be formed with baselines shorter than the critical one. Inthe following, we will use the pair with the better azimuth scan-ning pattern overlap. The properties of this data pair are sum-marized in Table I.

III. B URST-MODE SIGNAL AND SPECTRAL PROPERTIES

This section reviews the unique properties of burst-modedata. In Section III-A we recall some important basics of SARprocessing in order to lay a common ground concerning theterminology used in the paper.

A. SAR Signal Processing Preliminaries

Since SAR data acquisition and processing are linear oper-ations, it is sufficient to consider how a point-like scatterer isimaged. Let a (strip-map) SAR fly along a straight path withvelocity [m/s] and illuminate a point scatterer at a distance

[m] to the flight path. The position of the scatterer is furtherdetermined by the time [s], when SAR and scatterer have aminimum distance (point of closest approach or zero-Dopplerposition). Then and reference the scatterer position in acylindrical radar coordinate system (the third coordinate—thelook angle—is only accessible by interferometry).

The instantaneous distance of the SAR to the scatterer at az-imuth time [s] is

(1)

TABLE IRADARSAT DATASETS USED FORSCANSAR INTERFEROMETRY INTHIS

PAPER. WE DENOTE VARIATION FROM NEAR TO FAR RANGE BY “…”

SAR processing algorithms must consider the hyperbolic formof the range history, while for our derivations the quadraticTaylor series approximation of (1) is sufficient and will be usedthroughout the paper.

Let the SAR periodically transmit pulses of the form, where [s] denotes the fast (range) time

unique to each pulse with origin at the particular transmit eventand where [Hz] is the radar carrier frequency. In the case ofa chirped pulse, the function is a complex-valued chirp.All received pulses are replicas of the transmitted pulse delayedby the round-trip delay time , where [m/s]represents the velocity of light, with amplitude given by theradiometric radar equation. The only radiometric componentrelevant for this paper is the two-way azimuth amplitudeantenna pattern (scaled to the azimuth time coordinate).After coherent demodulation, i.e., removal of the radar carrierfrequency, the ensemble of received pulses is given by

(2)

This point response shows a range sweep characterized by, while and are the envelope

and the phase term of the azimuth chirp, respectively.SAR processing “convolves” the raw data with an appropriate

filter kernel such that every raw data point response gets fo-cused to a sharp point as narrow as the range and azimuth band-widths of the system allow. In fact, the required operation isnot a convolution in the strict sense, since the point responseto be inverted is range-variant [see (2)]. SAR focusing usuallystarts with range compression, i.e., the application of a simplerange-matched filter to compress the transmitted range chirp.After that operation, the raw data look like a short pulse, i.e., as

HOLZNER AND BAMLER: BURST-MODE AND SCANSAR INTERFEROMETRY 1919

Fig. 2. Raw data of a single-point scatterer recorded in strip-map (top) and in four-look burst-mode (bottom). To illustrate the oscillations in these complex-valuedsignals, only the real parts are shown. The range migration is exaggerated to illustrate the SAR response properties.

if a sinc-like envelope was transmitted. Fig. 2 (top) illustratessuch a range-compressed raw-data-point response. Clearly vis-ible is the range sweep or range cell migration effect due tothe range history in the envelope term , ren-dering the problem of SAR processing not only range-variantbut also two-dimensional. This means that SAR processing isnot readily separable into independent azimuth- and range-fil-tering steps. In order to achieve separability, most of the SARprocessing algorithms employ an operation called range cell mi-gration correction that straightens the range trajectory prior to apurely one-dimensional (1-D) range-variant azimuth filtering.

For the purpose of this paper, it is sufficient to know thathigh-quality strip-map SAR focusing algorithms exist (e.g., see[12] and [13] and references therein). We will introduce anddiscuss burst-mode processing algorithms that take maximumadvantage of existing strip-map algorithms as far as possible.Since burst-mode data differ from strip-map data only in theazimuth direction, we can simplify the equations and consideronly the azimuth direction of the SAR signal. For convenience,we will use the quadratic approximation of the range history of(1), the azimuth (Doppler) frequency modulation rate

FMHzs

(3)

and the Doppler frequency envelope generated by the azimuthantenna pattern

FM(4)

B. Strip-Map Signal

Now, let [Hz] be the (Doppler) frequency domain coordi-nate conjugate to azimuth time, PRF [Hz] the pulse repetitionfrequency (PRF), and [Hz] the processed Doppler band-

width of a strip-map (nonburst) SAR signal. Then, the syntheticaperture length is

FMs or PRF number of range lines

(5)The azimuth resolution of such a SAR system is (in time units)

FMs (6)

where is a factor close to unity that depends on the shape ofthe azimuth antenna pattern and on weighting functions usedby the SAR processor for side-lobe control. For simplicity, wewill use throughout the paper. The effective strip-mapazimuth raw data signal of a point scatterer located at isthen given by

FM FM (7)

with spectrum (stationary phase approximation)

FM(8)

Here, we have assumed zero-Doppler steering of the antennabeam and have discarded an inconsequential factor in the spec-trum. is the spectral envelope and reflects the two-wayDoppler antenna pattern (Section III-A). For mathematicalconvenience, we also include any weighting and band-lim-iting function applied during processing in . Within thestationary phase approximation, FM is the time-domainenvelope of the azimuth chirp. Its support is restricted to

.SAR processing removes the quadratic phase from the

spectrum of (8). Hence, the shape of the focused point targetresponse is the inverse Fourier transform of

(9)


Fig. 3. Train of raw-data bursts and its spectrum (magnitude) according to (14).

where corresponding small and capital letters denote theFourier transform pairs [e.g., ]. Since isband-limited to , has the resolution givenin (6).

C. Burst-Mode Signal

Consider now that the SAR system characterized above isoperated in the burst-mode such that bursts ofconsecutiveechoes are recorded every [s]. We will refer to

PRFs or number of range lines (10)

as the burst duration and to

s or PRF number of range lines (11)

as the burst cycle period. Note that in general is notinteger, since the SAR may be operated at a different PRFbetween two bursts [5]. This subtlety is not substantial forthe following signal description and will not be consideredfurther. Throughout the paper, we assume that the numberof range lines per burst is large enough for a sufficiently hightime-bandwidth product, such that we can use the stationary

phase assumption. The low time-bandwidth-product case hasbeen investigated in [14].

The number of bursts per synthetic aperture

(12)

is often called the number of looks. Fig. 2 illustrates the relation-ship between the strip-map and the burst-mode SAR raw data ofa single point scatterer. For the illustrations in Figs. 2, 3, 5, and9, we use .

The burst-mode raw-data-point scatterer response can bedescribed as

FM

rect FM (13)

with spectrum (within the stationary phase approximation)

rectFM

FM(14)


Fig. 4. Time-frequency diagram of focused multiple-burst RADARSATScanSAR data in a distributed target area (f : Doppler centroid frequency).

where and are the burst bandwidth and the spectralburst cycle period, respectively:

FM and FM (15)

Fig. 3 illustrates the relationship of the burst-mode signalto its spectrum. Each burst contributes an approximatelyrectangular-shaped band-pass spectrum at center frequency

FM , which depends on the location ofthe point scatterer and the burst number. We will neglectthe slight distortion of these rect-functions due to inthe following. Weighting of the bursts is often applied inthe processor for purposes of side-lobe control. In that case,the single-burst spectrum reflects the shape of the weightingfunction rather than a rect-function.

If each of the bursts is processed individually, its point re-sponse will be

FM (16)

with the inherent single-burst resolution of

FMs (17)

If the full-burst train is processed coherently, the spectral andtime-domain representations of the coherent point response willbe [5]

rectFM

(18)

(19)

Fig. 4 illustrates the time-varying property of the transfer func-tion (see also [6]) estimated from real data in a distributed targetarea (ScanSAR dataset 10 969, Table I). The energy of the burstdata is represented in the domain spanned by azimuth frequencyversus azimuth time. It is concentrated in areas that show thementioned azimuth dependency of the burst raw-data signal. Forone scatterer, the spectrum of the processed signal is of band-pass type with bandwidth FM centered around

FM [see (18)]. Thus, with varying scatterer

Fig. 5. Relationship of single-burst and burst train point responses. The smallSAR image patch shows the response of a corner reflector processed in the bursttrain fashion [5].

Fig. 6. Doppler centroid estimates^f (circles) as a function of range andfitted second-order polynomial (data: RADARSAT ScanSAR Narrow, orbit10 969).

position the entire Doppler bandwidth to isswept. The distance between the bursts in time coordinates is theburst period time , and in Doppler domain it is the frequencydifference .

The same point response as given in (19) will apply if theindividually focused bursts are summed coherently. Obviously,

can be interpreted as a train of full-aperture (strip-map)point responses with the single-burst point response

as an envelope (Fig. 5).

IV. PREPROCESSING

A. Doppler Centroid Estimation

Doppler centroid estimation of burst-mode data is more dif-ficult than in the strip-map case due to the particular shapeof the spectra. Spectra are distorted according to the azimuthintensity profile of the target. They require sufficiently largeestimation windows to average out the distortions. In our im-plementation, the bursts are packed densely in the azimuth di-rection—giving an uninterrupted sequence of range lines—andthen passed through a standard Doppler centroid estimator. Thisprocedure fits well into the pack-and-go algorithm describedlater.

Fig. 6 depicts the Doppler centroid estimates as a functionof range and the resulting second-order polynomial fits. Theestimates are based on range-compressed data. We have usedwindow sizes of 10 000 range lines by 150 range samples, i.e.,


Fig. 7. Range displacement of two azimuth looks of a point-like target asa function of PRF bands. The correct PRF band is�5 (data: RADARSATScanSAR, orbit 10 969).

45 and 53 estimation windows for beam 1 and beam 2, respec-tively. A quadratic fit to these estimates suggests a standard de-viation of the individual estimates of about 20 Hz for beam 1and 10 Hz for beam 2. After fitting the Doppler centroid, theerror reduces to about 3 Hz and is thus negligible. The interfer-ometric image pairs were processed to the average of the esti-mated Doppler centroid frequencies of the two datasets to avoidspectral decorrelation in azimuth.

B. PRF Ambiguity Resolution

RADARSAT is not yaw-steered like ERS or ENVISAT.Hence, Doppler centroid values exceedingPRF/2 are ex-pected, making PRF ambiguity resolution necessary. Wemodified the range look correlation technique [15] for thispurpose. Since ScanSAR scenes cover areas about 10 times aslarge as strip-map images, it is very likely that there are a fewpoint-like targets in the scene. These points can be detectedautomatically and used for range look correlation. Fig. 7 showsthe range cross-correlation offset of two looks of a point-liketarget. In this case, a safe determination of the PRF band waspossible using a single point. In less favorable cases, the resultsfrom several points must be averaged.

C. Azimuth Scanning Pattern Synchronization

Interferogram formation from burst-mode data requires thesynchronization of the burst patterns [azimuth scanning patternsynchronization (ASPS)] [6], [16]. The following procedurewas applied [16]. First a small azimuth portion of the masterand slave dataset is processed to a complex image. Thismeans that several bursts are processed and added coherentlyat the acquisition PRF (multiple coherent burst method; seeSection VI-C). Then (for each beam separately) a patch atmidrange is taken to determine the mutual azimuth shiftbetween the two images. We implemented the measurementby forming interferograms for different mutual shifts. The lagwith the maximum interferogram spectral peak is chosen. Thisallows for azimuth coregistration of the two burst-mode rawdatasets to an accuracy of at least one range line, i.e., sufficientalong-track referencing of the two azimuth-scanning patterns.

Fig. 8. Peak value of interferogram spectrum versus mutual azimuth shift forazimuth scanning-pattern synchronization.

Fig. 8 shows the variation of the spectral peak maximum asa function of mutual shift after synchronization of the scanningpatterns. The strong peak of the curve suggests sufficiently ac-curate ASPS. An obvious feature of the curve is its modulation,which is caused by the particular correlation properties of theburst-mode data (Section VI-D).

Once the mutual displacement of the two datasets has beendetermined, only the range lines that have counterparts in theother dataset’s on-cycle are kept for further processing; all theothers are discarded [16]. Table I lists the number of remainingburst lines and the corresponding burst bandwidth for the datasetunder investigation. Each burst has a bandwidth of about 95 Hz,resulting in an azimuth resolution on the order of 75 m. Thisresolution is better than the one of DTED-1 with sampling gridof about 100 m [17].

V. PHASE-PRESERVINGBURST-MODE PROCESSING

Several algorithms have been proposed for phase-preservingprocessing of burst-mode data [4]–[8], [18]. The basic ques-tion is: what does a complex burst-mode image product looklike? Two approaches are possible with implications for datavolume and further interferometric processing: single- and mul-tiple-burst image processing.

A. Single-Burst Images

In this option, the bursts are processed to individual compleximages. The complex burst-mode product is then the ensembleof these burst images. Each burst image has a resolution of

[see (17)] and covers an azimuth extent of .Allowing for the same azimuth oversampling factor as the rawdata, its sampling frequency will be

PRF PRF (20)

and it is sufficient to represent the burst image by

PRF PRF

range lines. Per synthetic aperture and subswath, wethus have burst images of range


Fig. 9. Pack-and-go range processing of a burst train. For illustration purposes, the raw data are assumed to be pre-range-compressed.

lines each, i.e., a data volume and processing effort ofPRF lines. The factor

(21)

is approximately the number of ScanSAR subswaths (or polar-ization modes in the ASAR case).

An inconvenience associated with individual burst images istheir azimuth-variant spectral bandpass characteristic (time-fre-quency diagram in Fig. 4). Although the signal is of bandpasstype in a small neighborhood of every azimuth position, thebandpass center frequency varies linearly with azimuth from

to and wraps around several times in thePRF Nyquist band of the burst image. This excludes the

use of standard coregistration and resampling algorithms forlater interferometric processing.

B. Multiple-Burst Images

An alternative complex burst-mode product would be thecoherent superposition of the burst images forming a singlefull-aperture image file. Such a representation facilitates datahandling and visual browsing, although it suffers from the“nasty” shape of the point response of Fig. 5. It has been shown,however, that a low-pass filter applied to the detected imagegives an image comparable to the one obtained by incoherentsuperposition of burst images [5]. The average spectral enve-lope of such a complex burst-mode image is equivalent to thatof a strip-map SAR signal. Hence, standard interferometricprocessing algorithms can be applied (Section VI). The priceto be paid is increased data volume. The full-aperture imagerequires the same PRF as the raw data and, hence, has a datavolume higher by a factor of than inthe aforementioned individual burst case. If the superposition

Fig. 10. Azimuth power spectral densities (normalized) of strip-map,single-burst, and multiple-burst interferograms. Stationary Gaussian scatteringand a linear phase ramp, i.e., a single fringe frequencyf , are assumed. Thetriangular shapes of the spectra apply for rect-shaped burst profiles. In case thebursts are weighted for side-lobe control, the triangular spectra are replacedby the autocorrelation of the weighting function. Note that the position of theDirac distribution is restricted to the support [�W ,W ] or [�W ,W ] forthe strip-map mode or burst-mode data, respectively.

image is to be formed from single-burst images, the data willhave to be oversampled from PRFto PRF before adding themcoherently.

For the following systematic overview of burst-modeprocessing strategies, we have conceptually separated the pro-cessing chain into range and azimuth processing. The first stepincludes range compression, range cell migration correction,


Fig. 11. Measured ACF of (a) ScanSAR single look, (b) multiple (N = 2), (c) comparison of both, and (d) histogram of 121 coregistration estimates.

and secondary range compression, leaving the second step as apurely 1-D azimuth compression (AC).

C. Range Processing

Within this section, we will focus on modifications of high-precision strip-map algorithms. Standard SAR processing al-gorithms [12] such as range-Doppler, wavenumber domain, orchirp scaling can be used for range processing of burst-modedata. Three approaches are possible.

1) Single-Burst Range Processing:Each burst of range linesis processed separately by a standard processor whose azimuthcompression has been disabled. This method is computation-ally efficient and has been adopted for use in the extended chirpscaling (ECS) algorithm [7] and in the ScanSAR extension ofour BSAR processor [19]. However, often the bursts must besupplemented by a number of range lines containing zeroes tothe next power of two.

2) Pack-and-Go Range Processing:Since during rangeprocessing (almost) no energy leakage in azimuth occurs,


Fig. 12. (a) Differential phase of the beam 1 and beam 2 interferogram in the overlap area after reduction of a phase offset of�3.11 rad. The range mean of thephase versus azimuth in (b) shows no systematic effects.

several bursts can be concatenated (with a few zero-lines as“safety margins” separating the different bursts) and processedsimultaneously by the standard SAR processor. After rangeprocessing, the bursts can be easily separated (Fig. 9). Thismethod is more efficient than single-burst processing, sincearbitrarily long FFTs can be used and zero padding (to meet thepower-of-two requirement) is negligible. Existing strip-mapSAR processors can be used like a black box, no matter howlarge their internally used azimuth processing blocks are.Also Doppler centroid estimation that uses large estimationwindows can be conveniently applied to this packed burstdataset (Section IV-A). We denote this new efficient algorithmby “pack-and-go.”

3) Coherent Full-Aperture Multiple-Burst Processing:Aless efficient algorithm for range and azimuth processing hasbeen proposed in [5]: the interburst pauses are filled with zeroessuch that the burst train resembles a standard strip-map-moderaw dataset where some range lines (the pause lines) have beendeliberately set to zero. The resulting coherent burst train (asdepicted in Fig. 9, top) is then processed in range and azimuthjust like a standard strip-mode raw dataset. Also, this algorithmcan use existing strip-map SAR processors conveniently at theexpense of reduced throughput. The resulting image is alreadysampled at the final sampling rate PRF, and oversampling ofburst images is not required.

D. Azimuth Processing

1) SPECAN: The classical SPECAN algorithm is compu-tationally efficient and can be made phase-preserving. A fo-cused burst image, however, exhibits a range-dependent azimuth

sample spacing known as fan-shape distortion and, hence, re-quires an additional interpolation step.

2) Azimuth-Variant Filtering: After a burst has been rangeprocessed, standard azimuth compression is applied. The re-sult is a coherent superposition of several shifted replicas ofthe desired burst image. A time-domain azimuth band-pass filterwhose center frequency varies with azimuth position is requiredfinally to separate the correct result from the aliased one [6].

3) ECS: The high-precision ECS algorithm [7] is a com-bination of chirp scaling for accurate range processing andSPECAN for azimuth processing. An azimuth scaling step isadded to equalize the frequency rates of the azimuth chirps.Due to this trick, 1) fan-shape distortion of the classicalSPECAN is avoided; 2) no interpolation is required; and 3) thefinal azimuth sampling grid can be chosen arbitrarily. This alsoallows for mosaicking the processed subswaths (possibly takenat different PRFs) to the complete ScanSAR image withoutfurther interpolation.

4) Time-Domain AC:Since the bursts are usually short (e.g.,100 range lines), a straightforward time-domain azimuth corre-lation is an alternative to frequency domain methods. Althoughnot as efficient, it is most easily implemented and allows fora free choice of the output sample spacing. We used this flex-ible method to process the RADARSAT ScanSAR data (Table I)used in this paper.

5) Coherent Multple-Burst Processing:If the bursts havebeen range processed according to the above-mentionedcoherent multiple-burst range processing technique (Sec-tion V-C3), standard azimuth compression can be applied to theburst train. The result is equivalent to a coherent superposition


Fig. 13. Phase and amplitude of the two-beam RADARSAT ScanSAR interferogram.


Fig. 14. Stability of the complex interferogram phase. The phase stability increases (brighter values) in the beam overlap area.

of the (correctly coregistered and oversampled) individual burstimages [5]. As a consequence, the azimuth impulse responsefunction exhibits the mentioned strong interference modulation(Fig. 5).

VI. BURST-MODE AND RADARSAT SCANSARINTERFEROMETRICPROCESSING

With the complex burst-mode images available, there aretwo options for further interferometric processing that willbe discussed in the following:single-burstandmultiple-burstinterferograms. We will derive the power spectral densities ofthe different types of burst interferograms. In order to simplifythe equations, we will again restrict ourselves to theazimuthdirection of the SAR signal. As a reference, we start with areview of spectral properties of strip-map interferograms.

A. Spectrum of Strip-Map Interferograms

We assume homogeneous targets, i.e., Gaussian stationaryscatterers with the following complex reflectivity functions[20]:

(22)

The functions , , and are independent zero-meanwhite circular complex Gaussian processes. We denote the co-herence between both signals as(temporal coherence) and apossible azimuth fringe frequency as. We assume sufficiently

small analysis windows to consider constant (azimuth spec-tral shift is discussed in Section VI-G). The cross correlation ofthe two processes and is given by

(23)

where is the Dirac distribution, and the asterisk “” de-notes complex conjugate.

The focused strip-map SAR data, (master) (slave),is obtained by convolution with the end-to-end system impulseresponse function from (9) by the following integral:

(24)

Moreover, we use the standard definition of the coherence of aninterferogram pixel :

(25)

Using (24) and Parseval’s theorem, we find for the coherence ofa strip-map signal [21]

(26)

the transfer functions are the Fourier transforms of theresponses .


In order to find the power spectral density (PSD) of thestrip-map interferogram, we calculate the autocorrelationfunction (ACF) of the interferogram and transform itinto Fourier domain. The interferogram ACF is obtained by

(27)

and by using the well-known Reed’s theorem [22], we find

(28)

with and being the ACFs of and, respectively. For better comparison of the individual

interferograms—strip-map, single, and multiple coherentburst—we define the following normalized version of theinterferogram ACF:

(29)

By inserting (28) and (25) as well as taking into account thenotation for the coherence in (26) we find

(30)

We see that the last term in the second line of (30) is directlyrelated to coherence magnitude and the interferometric phaseramp. Since the ACF of the interferogram only depends on(the time lag between the considered interferogram pixel) andsince the coherence magnitude in (26) is a constant, we considerthe interferogram process as wide-sense stationary. The Fouriertransform of (30) is then the normalized PSD of the strip-mapinterferogram

(31)

where “ ” is the correlation operator.In summary, the power spectrum of the interferogram is the

cross correlation of the power spectra of the individual channelsplus a Dirac distribution centered around the fringe frequencywith a weight equal to the square of the coherence of the twoSAR signals.

For example, consider two identical strip-map SAR systemswith a rectangular-shaped transfer function—perfectly coregis-tered SAR signals with the same Doppler centroid. Then, thecoherence of (26) is given by

tri (32)

Fig. 15. ScanSAR interferogram and DEM (Figs. 13–15) cover an area ofabout 110 km in azimuth and 350 km in range. It includes the city of LasVegas, the Colorado River, and the Grand Canyon (map: © www.expedia.com/dayily/home/default.asp).

This coherence can be maximized by processing, i.e., azimuthspectral shift filtering. The maximum obtainable value will bethe temporal coherence component.

Since the strip-map SAR impulse response is time-invariant,the PSD of the SAR signals is also rectangularly shaped (as-suming zero-Doppler steering):

rect (33)

Consequently, their correlation is a trifunction with bandwidth2 centered around zero frequency. Thus, we obtain the fol-lowing for the normalized power spectrum from (31) and using(32):

tri tri

(34)Fig. 10 illustrates this normalized PSD. Obviously, is theratio of the impulse integral of the spectral line at the centerfringe frequency and the integral of the extended trianglespectrum.

B. Single-Burst Interferograms

Given the individual complex burst images of the masterdataset and the corresponding ones of the slave data, smallinterferograms are generated from each pair of companionburst images. These are referred to as “burst interferograms”or “interferometric looks.” In a second step, the burst inter-ferograms are properly shifted in azimuth and are concatenatedor (in case of overlap) coherently added to form a full inter-ferogram of the subswath under consideration. With the usualScanSAR scenarios, several () of these burst interferogramsoverlap at every point of the final interferogram. This leadsto a phase noise reduction equivalent to-look processing.Since the variation of the bandpass center frequency withazimuth inherent to the complex burst images is the same inthe corresponding master and slave image, it cancels out in theburst interferogram. Therefore, the burst interferograms can be


Fig. 15. (Continued)Resulting ScanSAR DEM. The Grand Canyon is stretching from the middle to the right edge of the DEM. In the upper left part of the DEM,the plane (about 20 km�20 km) hosting the city of Las Vegas is visible. The region used for quality assessment (Fig. 16) is marked by a rectangle.


added without oversampling. The resolution of the final-look interferogram is .

In analogy to the derivation of the strip-map interferogramspectrum, we find the normalized PSD of multilook aver-aged single-burst interferograms (Fig. 10)

tri tri

(35)Note that due to the multilooking of the burst interferograms, theenergy of the extended triangle is reduced by a factor of ;hence, coherence or phase stability is larger than in the strip-mapcase.

C. Multiple-Burst Interferograms

For the multiple-burst interferometric processing, the burstimages from each beam are added coherently before interfero-gram formation as described above. This is done automatically,if the coherent multiple-burst range and azimuth processing al-gorithm has been used (Sections V-C3 and V-D5). The result isan image with bandwidth but having the oscillating azimuthimpulse response function of Fig. 5. From the two high-band-width images, we form the interferogram and obtain for the nor-malized PSD

tri

tri (36)

Subsequent azimuth averaging of adjacent interferogramsamples, i.e., low-pass filtering with a filter anddown-sampling, reduces the bandwidth and sampling fre-quency to that of a single-burst interferogram and removes theoscillations in the point response function. The normalizedPSD of the filtered multiple-burst interferogram is then

(37)

We see that for a filter that picks our the center triangle spectrum,e.g., rect , the spectrum of the low-passfiltered multiple-burst interferogram (37) and the spectrum ofthe single-burst multilook interferogram (35) are equivalent.

As already mentioned, the intermediate increase in bandwidthrequires redundant sampling of each burst image and increasesthe data volume input to the interferometric processor. On theother hand, at the higher spectral resolution, the average spec-trum of this data resembles strip-map spectra, and, therefore,conventional coregistration and resampling techniques can beapplied (see Section VI-E).

D. Estimation of Coregistration Parameters

Coregistration parameters between the master and the slaveimages are often estimated by cross-correlating homologousimage chips. The accuracy of mutual azimuth shift estimatesis proportional to the resolution of the images [23]. In thecase of single-burst images, this means , where thefactor comes in, since at any point independent shiftestimates can be derived.

For multiple-burst images, estimation of the mutual azimuthshift requires some care: due to the modulation of the impulseresponse function, multiple local correlation maxima will occur(Figs. 8 and 11). Also, coherence as a function of azimuthmisregistration exhibits multiple maxima.

If we can find the global maximum, the inherent resolutionbecomes the determining factor for coregistration

shift estimation. Since , the expected estima-tion accuracy will be better by a factor of compared tothe single-burst interferogram case.

An elegant method for mutual azimuth shift estimation basedon the so-called spectral diversity algorithm is proposed in [24].It exploits the fact that the phase offset between the interfero-grams formed from different looks is proportional to the azimuthregistration error. Since in burst-mode interferometry the looksare already available, spectral diversity is easily applicable.

E. Interpolation for Image Coregistration

Resampling of the slave image for coregistration with re-spect to the master image requires interpolation. A multiple-burst image can be interpolated in the same way as strip-mapdata because its average spectral density and samplingfrequency PRF are the same as in the strip-map case.

The azimuth interpolation of single-burst images, however,requires some care, since the time variance of the burst imagespectrum must be considered. Standard interferometric pro-cessing systems assume a constant azimuth center frequency(Doppler centroid), while in burst-mode data the look centerfrequency varies linearly with azimuth. Therefore, an az-imuth-variant interpolation kernel must be used.

An efficient interpolator for single-burst images is derived inthe following. According to (16), the response of a pointscatterer at has a linear phase whose slope is proportionalto . To a first approximation, this looks as if a gross quadraticphase were superimposed on the burst image. If we compensatefor that quadratic phase, the spectra will be shifted toward zerofrequency and become azimuth-invariant. Without loss of gen-erality, we may assume that we process burst . Then

FM(38)

As a first step, we multiply the burst image by FM ,resulting in

FM

FM FM

(39)

Now the linear phase in the point response is gone, and itsspectrum is of low-pass type. The residual quadratic phase

FM , however, broadens the spectrum. Usingas the equivalent width of the point response, the

bandwidth increase is

FM(40)

This increase of spectral width must be considered whenchoosing the sampling frequency PRFof the burst image.In the RADARSAT example presented in Section VII


(FM 2000 Hz/s), the burst bandwidth is in the orderof 100 Hz and, hence, increases by 20 Hz.

Once we have down-modulated the azimuth-variant spectraby the quadratic phase multiply, we can apply a standardlow-pass interpolator to map the time coordinate to the oneof the master image: , which is more or less a simpleshift. Finally, we compensate for the gross quadratic phase(properly mapped to the master coordinates) by multiplicationwith FM .

The problem of azimuth-variant band-pass interpolation canbe avoided also, if the slave image is processed (or reprocessed)such that the necessary shift operation is performed in theprocessor. This is done most conveniently, e.g., by introducinglinear phase functions in the range-Doppler domain. The ECSalgorithm (see Section V-D3 and [7]) further allows a straight-forward implementation of a linear azimuth scaling factor byproperly manipulating the chirp functions involved [24].

F. Variation of Coherence With Varying SNR

We do not correct for scalloping, and we obtain a varyingsignal-to-noise ratio (SNR) in the azimuth direction [25]. Thisvariation might also be seen in the phase or coherence images,since the coherence depends on the SNR of the SAR images[21], [26]. With our dataset, however, this effect is negligible,since temporal decorrelation dominates.

G. Azimuth Spectral Shift

A mutual azimuth spectral shift in SAR images is a conse-quence of an azimuth terrain angle or converging orbits [6], [20].The azimuth fringe frequency is given by the following equation[6]:

Hz (41)

with (the perpendicular baseline), (look angle of theradar), and (azimuth terrain angle). For RADARSAT ata look angle of 40 and a perpendicular baseline of 100 m,we find that the terrain angle must be as high as 22for afringe frequency in the order of a tenth of the burst bandwidth.Hence, for our dataset this effect is negligible. However, forthe very small bandwidth of the ENVISAT Global Monitoringmode, the azimuth spectral shifts are significant and should becompensated [6].

H. Beam Alignment

A ScanSAR dataset can comprise 2–5 beams as plannedfor ENVISAT/ASAR operation [3]. The overlap area can beused for calibration of the ScanSAR data as described in [27],and in case of swath interferogram combination, it can be usedfor beam coregistration verification and mutual phase offsetdetermination.

After careful ScanSAR dataset coregistration and interfer-ogram formation, there are still tiny misregistrations (in theorder of 1/100 sample) that lead, together with a high Dopplercentroid, to slowly varying phase errors. These are usuallyaccounted for by suitable orbit trimming. However, the misreg-istrations lead to a mutual phase offset in the overlap area thathas to be measured and compensated before beam combination.

Note that orbit trimming accounts for both beams, while theresidual misregistrations may be different in the individualbeams. In addition, the range fringe frequency will introduce aphase offset in the overlap area due to mutual misregistrationof the beam interferograms [28]. The mentioned phase offseteffects can be summarized by the following equations:

PRFaz az az (42)

and

RSFrg rg rg (43)

In (42) and (43), the spectral center frequencies are the Dopplercentroid frequency and the center of the range spectrum

(range spectrum off-set frequency [see(45)]. They trans-form the residual misregistrationsaz ( az ) and rg( rg ) for beam 1 (beam 2) into a phase error, e.g., see [29].The other contributions from the fringe frequencies and

occur due to the swath interferogram misalignmentsrgand az in range and azimuth, respectively.

For the dataset used, by far the most significant contribu-tion to phase error is azimuth misalignment of the ScanSARimages, since the Doppler centroid is very high, PRF6.0 cycles/sample. By contrast, the gross azimuth fringe fre-quency (converging orbits) impact is negligible (10 cycles/sample). The range frequencies are in the order of0.2 cycles/sample.

To measure beam alignment, we used multiple coherent burstimages of either dataset (Table I). Since the spectra of the com-plex beam signals are nonoverlapping, and hence uncorrelated,we measured the shifts on detected images.

For our particular RADARSAT dataset, we used severalhundred patches and found that the beams are well aligned inazimuth. However, a shift of 0.15 0.02 samples is present inthe range direction. After resampling the beam 2 interferogram,we measured a phase offset of3.11 rad (with a standarddeviation of 1.47 rad) and subtracted it from the beam 2interferogram before combining the beam interferograms. Theinterferogram pixels were added coherently in the overlaparea. The differential phase of the swath interferograms in theoverlap area (Fig. 12) shows no systematic effects such astrends or structure from topography originating from a mutualmisregistration. This implies that no azimuth-dependent intra-beam coregistration phase errors show up in the differentialphase, and it shows that interbeam coregistration is sufficientlyaccurate, so that no topographic artifacts are visible.

Finally, we obtained a full-range interferogram by beam com-bination. The interferogram phase and amplitude are illustratedin Fig. 13. It covers an impressive area of approximately 110 kmin azimuth and 350 km in range.

After burst image mosaicking, coherence information [see(25)] is not accessible anymore. The following phase stabilitymeasure was used instead:

(44)


Fig. 16. Comparison of different DEMs (GLOBE, DTED-1, DTED-2, ScanSAR) for the area (35 km� 35 km) marked in Fig. 15.

The result for the combined interferogram is shown in Fig. 14.As expected, we observe a reduction of phase variance inthe overlap area. This is further evidence that remainingtiny intrabeam azimuth misregistrations do not result inazimuth-dependent phase distortions in the individual beaminterferograms.

I. Range Spectrum Shift of RADARSAT Data

Since RADARSAT is not operated in yaw steering mode, highsquint angles occur [9], [11]. High squint angles result in anoff-zero range spectrum center frequency according to the fol-lowing equation [11], [28]–[30]:

(45)

.This offset varies considerably from near range to far range,

from 0.95 MHz to 3.8 MHz, due to the high Doppler cen-troid variation (Fig. 6). Since the bandwidth of the dataset isonly 11.58 MHz, the center frequency of the spectrum is sig-nificantly shifted away from zero and complicates spectral shiftfiltering and resampling.

J. Orbit Correction

RADARSAT orbit data have very low accuracy; therefore,orbit correction and baseline estimation become necessary [11],[31]. The along- and circular across-track accuracies are givenin [31] as 100 m and 20 m, respectively.

Orbit correction was split into two tasks. First, the primarychannel orbit was corrected for shifts in slant range and azimuth

direction. Second, the baseline lengthand angle were opti-mized, and the secondary channel orbit was corrected.

The timing shift for the primary channel orbit was estimatedusing an interferogram simulated from a high-quality DEM(DTED-2, 25-m sampling grid). For this task, usually groundcontrol points such as corner reflectors or transponders areused. However, for the considered region, no ground controlpoints were available. After maximizing the peak of the differ-ential interferogram spectrum (measured – simulated) in roughterrain, i.e., minimizing the fine structures on the differentialphase, we found a shift of 310 ns in range and 4.87 ms inazimuth.

Baseline can be estimated by minimizing the difference be-tween a simulated and the measured interferogram. In order toavoid phase unwrapping in advance of baseline optimization(low coherence, Grand Canyon, and mountainous areas), weproceeded as follows. We simulated the interferometric phasefor several baseline length and angle settingsand in theneighborhood of the initial interferometer configuration167 m and 159 . For this task, we used a DTED-2 DEM(25-m sampling grid), since we did not want to blame ScanSARInterferometry for the bad orbit of RADARSAT. We generateda matrix of differential or flattened ScanSAR interferogramsand considered the corrected baseline optimum when we ob-tain an overall flat phase. For this kind of baseline optimiza-tion, the wide range extent of the ScanSAR data is favorable,since it limits the space of possible correction parameter vectors( , ) that correct for the smooth phasefunction on the difference phase. We obtained

4.22 m, 0.0075 radfor the baseline error estimate.


(d)

Fig. 17. Histograms of the height differences (a) GLOBE—DTED-2,(b) DTED-1—DTED-2, and (c) ScanSAR—DTED-2.

VII. SCANSAR DEM EXAMPLE

In this section, we describe the last steps of the processingof the ScanSAR interferometric data, the resulting DEM, and acomparison to other available DEMs of the test site.

After baseline estimation (Section VI-J) we subtract aninterferogram simulated from the GLOBE DEM prior to phasenoise smoothing and phase unwrapping. For phase filtering,we use a Gaussian low-pass filter with a bandwidth that wematched to the fringe frequency distribution of a representativearea in the interferogram. This improves the unwrappingperformance of the applied minimum-cost-flow algorithm[32]. Fig. 15 shows the final ScanSAR 75–m resolution DEM[33]. Through the combination with GLOBE, the final resultcontains also information of the coarse DEM in low coherenceareas (dark in Fig. 14). These areas are restricted to extremeslopes and severe temporal decorrelation.

The DEM is compared to a high-quality DTED-2 (25-mgrid). In flat areas, we find a height standard deviation ofabout 3–5 m. This result matches well the expectations fromcoherence, which varies from 0.72–0.85 in such regions.

Fig. 16 shows a comparison of the ScanSAR DEM with sev-eral other DEMs. The area for comparison has an extent of35 km 35 km and altitudes ranging from 640–1860 m. Theother DEMs are GLOBE, DTED-1, and DTED-2, available atsampling grids of 1 km, 100 m, and 25 m, respectively. TheScanSAR DEM in the lower right corner certainly improves theGLOBE and even adds detail to the DTED-1. It is remarkablehow distorted the DTED-1 is for the chosen area around LasVegas. The DTED-2 serves as the reference in this comparison.

The histograms of the height differences of the GLOBE,DTED-1, and ScanSAR DEM with respect to the referenceDTED-2 are shown in Fig. 17.

The 1-km GLOBE DEM appears to perform well, since thechosen area contains sufficiently large patches of flat terrain.Also, a visual comparison suggests that the GLOBE DEM ofthis area had probably been produced from the DTED-2 data.The band-like artifacts in DTED-1 result in a bimodal error his-togram. Besides that, the error of the ScanSAR DEM is muchless than that of DTED-1, and its visually perceptible resolutionis much better and resembles more DTED-2 than DTED-1.

VIII. C ONCLUSION

ScanSAR will be increasingly used in future spaceborne SARsystems. ScanSAR interferometry is a complex task because1) the properties of the involved signals are quite different fromthose of conventional SARs and 2) some of the approximationsaccepted for standard InSAR processing have to be revised dueto the large swaths covered by ScanSAR.

This paper gives a complete overview of the interferometricproperties of ScanSAR signals and their consequences for pro-cessing. The different processing algorithms we have introducedand discussed should allow the reader to choose the one that ismost efficiently integrated into an existing strip-map InSARprocessing chain. The step-by-step recipe-like explanation ofprocessing a RADARSAT ScanSAR data pair to a large DEMincluded descriptions of many additional precautions and tricksthat have to be considered in ScanSAR interferometry. It wasshown that DEMs of at least DTED-1 quality can be generated byRADARSAT repeat-pass ScanSAR interferometry in the case ofmoderate terrain relief andsufficientlyhigh temporal coherence.

ACKNOWLEDGMENT

The authors would like to thank B. Schättler, H. Breit,U. Steinbrecher, A. Moreira, and J. Mittermayer (all of DLR)


for their contributions to the ScanSAR processing software;M. Eineder (DLR) for helpful technical discussions and formaking available an efficient interferometric phase simulator;W. Knöpfle (DLR) for mosaicking 550 DTED-2 DEM tilesand for his support with geocoding tools; and I. Cumming(University of British Columbia, Vancouver) for sharing hisexperience and advice.

REFERENCES

[1] K. Tomiyasu, “Conceptual performance of a satellite borne, wide swathsynthetic aperture radar,”IEEE Trans. Geosci. Remote Sensing, vol.GE-19, pp. 108–116, 1981.

[2] R. K. Moore, J. P. Claassen, and Y. H. Lin, “Scanning spaceborne syn-thetic aperture radar with integrated radiometer,”IEEE Trans. Aerosp.Electron. Syst., vol. AES-17, pp. 410–420, 1981.

[3] S. Karnevi, E. Dean, D. J. Q. Carter, and S. S. Hartley, “Envisat’s ad-vanced synthetic aperture radar: ASAR,”ESA Bull., vol. 76, pp. 30–35.

[4] I. G. Cumming, Y. Guo, and F. Wong, “Analysis and precision pro-cessing of RADARSAT ScanSAR data,” inProc. Geomatics in the Eraof RADARSAT, Ottawa, ON, Canada, 1997.

[5] R. Bamler and M. Eineder, “ScanSAR processing using standard highprecision SAR algorithms,”IEEE Trans. Geosci. Remote Sensing, vol.34, pp. 212–218, Jan. 1996.

[6] A. M. Guarnieri and C. Prati, “ScanSAR focusing and interferometry,”IEEE Trans. Geosci. Remote Sensing, vol. 34, pp. 1029–1038, July 1996.

[7] A. Moreira, J. Mittermayer, and R. Scheiber, “Extended chirp scalingalgorithm for air- and spaceborne SAR data processing in stripmap andScanSAR image modes,”IEEE Trans. Geosci. Remote Sensing, vol. 34,pp. 1123–1136, Sept. 1996.

[8] R. Lanari, S. Hensley, and P. A. Rosen, “Chirpz-transform basedSPECAN approach for phase-preserving ScanSAR image generation,”Proc. Inst. Elect. Eng., Radar, Sonar Navig., vol. 145, pp. 254–261,Oct. 1998.

[9] R. K. Raney, A. P. Luscombe, E. J. Langham, and S. Ahmed,“RADARSAT,” Proc. IEEE, vol. 79, pp. 839–849, June 1991.

[10] RSI. (1999) RADARSAT Data User Guide. RADARSAT International,Richmond, BC, Canada. [Online]http://www.rsi.ca/resources/educa-tion/handbook.htm.

[11] D. Geudtner, P. W. Vachon, K. E. Mattar, B. Schättler, and A. L. Gray,“Interferometric analysis of RADARSAT strip-map mode data,”Can. J.Remote Sens., vol. 27, pp. 95–108.

[12] R. Bamler, “A comparison of range-Doppler and wavenumber domainSAR focusing algorithms,”IEEE Trans. Geosci. Remote Sensing, vol.30, pp. 706–713, July 1992.

[13] R. K. Raney, H. Runge, R. Bamler, I. G. Cumming, and F. H. Wong,“Precision SAR processing using chirp scaling,”IEEE Trans. Geosci.Remote Sensing, vol. 32, pp. 786–799, July 1994.

[14] A. M. Guarnieri and P. Guccione, “Optimal “focusing” of low resolutionScanSAR,”IEEE Trans. Geosci. Remote Sensing, vol. 39, pp. 672–683,Mar. 2001.

[15] I. G. Cumming, P. F. Kavanagh, and M. R. Ito, “Resolving the Dopplerambiguity for spaceborne synthetic aperture radar,” inProc. IGARSS,Zürich, Switzerland, 1986, pp. 1639–1643.

[16] R. Bamler, D. Geudtner, B. Schättler, U. Steinbrecher, J. Holzner, J. Mit-termayer, H. Breit, and A. Moreira, “RADARSAT SAR interferometryusing standard, fine, and ScanSAR modes,” inProc. Radarsat ADROFinal Symp., Montreal, QC, Canada, 1998.

[17] U.S. Geological Survey,Digital Elevation Models, Data UsersGuide. Reston, VA: U.S. Geological Survey, 1993.

[18] I. G. Cumming, Y. Guo, and F. Wong, “A comparison of phase-pre-serving algorithms for burst-mode SAR data processing,” inProc.IGARSS, Singapore, 1997, pp. 731–733.

[19] H. Breit, B. Schättler, and U. Steinbrecher, “A high-precision worksta-tion-based chirp scaling SAR processor,” inProc. IGARSS, Singapore,1997.

[20] H. A. Zebker and J. Villasenor, “Decorrelation in interferometric radarechoes,”IEEE Trans. Geosci. Remote Sensing, vol. 30, pp. 950–959,Sept. 1992.

[21] R. Bamler and P. Hartl, “Synthetic aperture radar interferometry,”Inv.Prob., pp. R1–R54, 1998.

[22] I. S. Reed, “On a moment theorem for complex Gaussian processes,”IRE Trans. Inform. Theory, vol. IT-8, pp. 194–195, Apr. 1962.

[23] R. Bamler, “Interferometric stereo radargrammetry: Absolute heightdetermination from ERS-ENVISAT interferograms,” inProc. IGARSS,Honolulu, HI, 2000.

[24] R. Scheiber and A. Moreira, “Co-registration of interferometric SARimages using spectral diversity,”IEEE Trans. Geosci. Remote Sensing,vol. 38, pp. 2179–2191, Sept. 2000.

[25] R. Bamler, “Optimum look weighting for burst-mode and ScanSAR pro-cessing,”IEEE Trans. Geosci. Remote Sensing, vol. 33, pp. 722–725,May 1995.

[26] H. A. Zebker and R. M. Goldstein, “Topographic mapping from inter-ferometric synthetic aperture radar observations,”J. Geophys. Res., vol.91, pp. 4993–4999.

[27] A. P. Luscombe, “Using the overlap regions to improve ScanSAR cal-ibration,” in Proc. CEOS SAR Calibration Workshop, Noordwijk, TheNetherlands, 1993, pp. 341–346.

[28] R. Bamler and J. Holzner, “Processing techniques for repeat-passScanSAR interferometry,” inProc. EUSAR 2000, Munich, Germany,2000, pp. 719–722.

[29] M. Bara, R. Scheiber, A. Broquetas, and A. Moreira, “InterferometricSAR signal analysis in the presence of squint,”IEEE Trans. Geosci.Remote Sensing, vol. 38, pp. 2164–2178, Sept. 2000.

[30] C. Prati, F. Rocca, A. M. Guarnieri, and E. Damonti, “Seismic migrationfor SAR focusing: Interferometrical applications,”IEEE Trans. Geosci.Remote Sensing, vol. 28, pp. 627–640, July 1990.

[31] H. Rufenacht, R. J. Proulx, and P. J. Cefola, “Improvement ofRADARSAT image localization,” inProc. Geomatics in the Era ofRADARSAT, Ottawa, ON, Canada, 1997.

[32] M. Costantini, “A novel phase unwrapping method based on networkprogramming,” IEEE Trans. Geosci. Remote Sensing, vol. 36, pp.813–821, May 1998.

[33] A. Roth, “GEMOS-A system for the geocoding and mosaicking of inter-ferometric digital elevation models,” inProc. IGARSS, Hamburg, Ger-many, 1999.

Jürgen Holzner (S’02) was born on August 9, 1973.He received the Diploma degree in data and infor-mation engineering from the Georg-Simon-Ohm-Fachhochschule University of Applied Sciences,Nuremberg, Germany, in 1997. He is currentlypursuing the Ph.D. degree at the University ofEdinburgh, Edinburgh, U.K.

In 1997, he joined the German Aerospace Centre(DLR), Oberpfaffenhofen. He has worked since thenon processing techniques for SAR interferometry.The main projects he worked on included differ-

ential interferometry for subsidence monitoring, RADARSAT-1 ScanSARinterferometry, evaluation of SAR interferometry-derived digital elevationmodels (DEMs), and contributions to the SRTM interferometric processor.His current research interests are in ENVISAT IASAR alternating polarizationmode interferometry, baseline optimization, DEM evaluation, and comparisonusing multiresolution methods, as well as ScanSAR processing algorithms andspectral estimation algorithms.

Richard Bamler (SM’00) was born on February 8,1955. He received the Diploma degree in electricalengineering, the Eng. Dr. degree, and the “habilita-tion” in the field of signal and systems theory in 1980,1986, and 1988, respectively, from the Technical Uni-versity of Munich, Munich, Germany.

He was with the Technical University of Munichduring 1981 and 1989 working on optical signalprocessing, holography, wave propagation, andtomography. He joined the German AerospaceCenter (DLR), Oberpfaffenhofen, in 1989, where he

is currently Director of the Remote Sensing Technology Institute. Since then heand his group have been working on SAR signal processing algorithms (ERS,SIR-C/X-SAR, RADARSAT, SRTM, ASAR), SAR calibration and productvalidation, SAR interferometry, phase unwrapping, oceanography, estimationtheory, model-based inversion methods, parameter retrieval methods foratmospheric spectrometers (GOME, SCIAMACHY, MIPAS), content-basedimage query systems, and data mining for remote sensing archives. In early1994, he was a Visiting Scientist at the Jet Propulsion Laboratory, Pasadena,CA, in preparation of the SIC-C/X-SAR missions, where he worked onalgorithms for the SIR-C ScanSAR, along-track interferometry, and azimuthtracking modes. His current research interests are in algorithms for optimuminformation extraction from remote sensing data with emphasis on SAR. He isthe author of a textbook on multidimensional linear systems.

burst-mode and scansar interferometry - geoscience and …€¦ · possible to map 80% of the...

Documents