GEOS 657 – MICROWAVE REMOTE SENSING
GRADUATE-LEVEL COURSE AT THE UNIVERSITY OF ALASKA FAIRBANKS
Lecturer:
Franz J Meyer, Geophysical Institute, University of Alaska Fairbanks, Fairbanks; fjmeyer@[Link]
Lecture 9: Geometric & Radiometric Properties of SAR Data; Geocoding,
Geometric & Radiometric Correction
UAF CourseFranz
GEOS 657
J Meyer, UAF
GEOS 657 Microwave RS - 1
Problems to Solve when Trying to Use SAR Data for Geophysical Analysis
• Problem 1: SAR data suffers from geometric distortions owing to the side-looking observation geometry
• Solution: We will identify the distortions and look for ways to remove them
• Problem 2: SAR images appear very noisy, making interpretation harder
• Solution: We will describe the noise and talk about filtering methods
• Problem 3: SAR data is in the acquisition coordinate system defined by the Azimuth & Slant Range directions
• Solution: We will talk about Geocoding
Franz J Meyer, UAF
GEOS 657 Microwave RS - 2
Example of a Geometric and Radiometric Distortions in SAR Imagery
• Evident Geometric Distortion:
• One side of the mountain seems shorter than the other
3
• Evident Radiometric Distortion:
• One side of the mountain seems much brighter
McKinley National
Park Airport
ALOS PALSAR (L-band) data © JAXA (2007)
Franz J Meyer, UAF
GEOS 657 Microwave RS - 3
GEOMETRIC PROPERTIES OF SAR IMAGERY
Franz J Meyer, UAF
GEOS 657 Microwave RS - 4
Geometric Distortions are Caused by the Slant Observation Geometry of
SAR Systems
x
x
For SAR images processed to ‘zero-Doppler‘ geometry: y
z R
where R y0 y 2 H z 2
H R
R const
z
R
x
y0 y
Franz J Meyer, UAF
GEOS 657 Microwave RS - 5
Three Types of Geometric Distortions Occur As a Consequence of
Oblique Look Angle
Foreshortening Layover Shadow
• Sensor-facing slope • Mountain top overlain on • Area behind mountain
forshortened in image ground ahead of mountain cannot be seen by sensor
• Foreshortening effects • Layover effects decrease • Shadow effects increase with
decrease with increasing with increasing look angle increasing look angle
look angle
Franz J Meyer, UAF
GEOS 657 Microwave RS - 6
Geometric Distortions of SAR Images
1. Foreshortening
range
A
foreshortening
B
C
A B C
ground range
ERS-1
23 deg
data © ESA
Franz J Meyer, UAF
GEOS 657 Microwave RS - 7
Geometric Distortions of SAR Images
2. Layover
range
lay-over
A C
B D
A B C D
ERS-1
23 deg
data © ESA
Franz J Meyer, UAF
GEOS 657 Microwave RS - 8
Geometric Distortions of SAR Images
3. Shadow
radar shadow
azimuth
range
SRTM/X-SAR
54 deg
Franz J Meyer, UAF
GEOS 657 Microwave RS - 9
Example of a Geometric and Radiometric Distortions in SAR Imagery
• Questions:
• Which direction is range and which is azimuth?
• Where (if at all) do you see foreshortening? 3
• Where (if at all) do you see layover?
• Where (if at all) do you see radar shadow?
McKinley National
Park Airport
ALOS PALSAR (L-band) data © JAXA (2007)
Look angle of image: ~ 35°
Franz J Meyer, UAF
GEOS 657 Microwave RS - 10
Courtesy A. Thiele, KIT & Fraunhofer IOSB, DE
Analysis of Complex Structures Contact author @:
[Link]
Simulation of SAR Signatures (Dr. H. Hammer, S. Kuny)
Range
TerraSAR-X image (Giza, Egypt)
Franz J Meyer, UAF
GEOS 657 Microwave RS - 11
Courtesy A. Thiele, KIT & Fraunhofer IOSB, DE
Analysis of Complex Structures Contact author @:
[Link]
Simulation of SAR Signatures (Dr. H. Hammer, S. Kuny)
Range
Range
incidence angle ≈ pyramid slope
TerraSAR-X image (Giza, Egypt)
Optical images © Berthold Werner, Jon Bodsworth (Wikipedia)
Franz J Meyer, UAF
GEOS 657 Microwave RS - 12
Courtesy A. Thiele, KIT & Fraunhofer IOSB, DE
Analysis of Complex Structures Contact author @:
[Link]
Simulation of SAR Signatures (Dr. H. Hammer, S. Kuny)
Range stepped pyramid
Range
TerraSAR-X image (Giza, Egypt)
Franz J Meyer, UAF
GEOS 657 Microwave RS - 13
Another Cool Example of Geometric Distortions
Franz J Meyer, UAF
GEOS 657 Microwave RS - 14
RADIOMETRIC PROPERTIES OF SAR IMAGERY
Franz J Meyer, UAF
GEOS 657 Microwave RS - 15
Think – Pair – Share
• Geometric Distortions can be challenging…
– Try to explain this scene. What object are you seeing? What is the
observation direction in this scene. What causes all the different bright
spots?
• Speckle is not noise, he says …
– I am claiming that Speckle is not noise in the traditional sense but rather a
consequence of the imaging process resulting from the big two principles in
SAR, Interference & Coherence. Try to discuss how interference effects in
coherent imaging systems may naturally cause speckle?
– Can you think of other sensors (other than SAR) that also may suffer from
Speckle?
Franz J Meyer, UAF
GEOS 657 Microwave RS - 16
SAR Images Often Appear a Bit Noisy
• Do you see the noise?
• This noise is caused “Speckle” and is an
inherent property of all coherent imaging
systems
• Technically speaking, it is not noise but an
interference pattern
[Link]
view-titan-despeckle-it
Franz J Meyer, UAF
GEOS 657 Microwave RS - 17
Coherent Waves and Speckle
Imu t
A
0 p/2 p 3/2p 2p Phase (t)
Reu t
0 90º 180º 270º 360º
Franz J Meyer, UAF
GEOS 657 Microwave RS - 18
Coherent Waves and Speckle
Summing contributions
within a pixel
Franz J Meyer, UAF
GEOS 657 Microwave RS - 19
Speckle
• Random positive and negative interference of wave contributions from the many individual scatterers
within one resolution cell
varying brightness from pixel to pixel even
for constant 𝜎 0
granular appearance
Franz J Meyer, UAF
GEOS 657 Microwave RS - 20
Speckle Statistics
amplitude power
quantity uI2 uQ2 I uI2 uQ2
probability
Rayleigh exponential
distribution
Important:
1. Noise level depends on radar cross
section 𝜎 0 → higher 𝜎 0 causes more
noise!
2. Noise is highly non-Gaussian!
1 I
pdfI I 0 0 exp 0
Franz J Meyer, UAF
GEOS 657 Microwave RS - 21
Speckle Statistics
amplitude power
quantity uI2 uQ2 I uI2 uQ2
probability
Rayleigh exponential
distribution
Important:
1. Noise level depends on radar cross
section 𝜎 0 → higher 𝜎 0 causes more
noise!
2. Noise is highly non-Gaussian!
1 I
pdfI I 0 0 exp 0
3. Averaging of 𝑁𝐿 independent samples
(looks) more Gaussian
I NL 1 NL L
N
I NL
pdfI I , NL
0
exp 0
NL 0 NL
Franz J Meyer, UAF
GEOS 657 Microwave RS - 22
Speckle Example
Time Series of SAR Images
incoherent average of 70 ERS SAR images individual images (9 years)
• Right image: shows how speckle can vary over time and in space
• Left image: shows that on average, the backscatter from an area is equal to its radar cross section 𝜎 0
Franz J Meyer, UAF
GEOS 657 Microwave RS - 23
Speckle Reduction
• SPECKLE is a scattering phenomenon and not noise. However, speckle can be modeled as multiplicative noise
for distributed targets (Lee, IGARSS-98)
• Speckle “masks” underlying image
• Speckle filtering:
– GOAL: Reduction of the speckle noise without sacrificing information content (including the spatial resolution)
– PRINCIPLE: Select homogeneous neighboring pixels and then average
• Simplest form of speckle reduction: averaging of adjacent pixels (box filter) or multi-looking loss of
resolution
• More complex models (try to limit resolution degradation)
Franz J Meyer, UAF
GEOS 657 Microwave RS - 24
Speckle Reduction
Example
Original
4-look 5x5 Median
amplitude
5x5 Boxcar Lee refined
(7x7)
Franz J Meyer, UAF
25
GEOS 657 Microwave RS - 25
Example for Bayesian Speckle Reduction
Original SAR Image Speckle Filtered
SAR data © AeroSensing GmbH Bayesian Algorithm
Franz J Meyer, UAF
GEOS 657 Microwave RS - 26
SAR Handbook: Comprehensive Methodologies for Forest Monitoring and Biomass
GEOS 657 Microwave RS - 27
Franz J Meyer, UAF
Estimation. Eds. Flores, A., Herndon, K., Thapa, R., Cherrington, E. NASA. 2019. DOI:
10.25966/hbm1-ej07
Selected Speckle Filters
GEOCODING, GEOMETRIC & RADIOMETRIC CORRECTION
Franz J Meyer, UAF
GEOS 657 Microwave RS - 28
Definitions
• Geocoding
– geometric transformation of an image into a cartographic map projection
• Georeferencing
– relating image coordinates to map coordinates by defining control points (usually image corners); image remains in
native image geometry
Franz J Meyer, UAF
GEOS 657 Microwave RS - 29
Geocoding Using a Sensor Model
• Sensor model
– sensor specific
– analytical reconstruction of image formation using orbit
and sensor parameters
– Digital Elevation Model (DEM) required to correct
image globally
Franz J Meyer, UAF
GEOS 657 Microwave RS - 30
Geocoding Steps
1. Relation between image coordinates and geographic coordinates using image and sensor geometry and
DEM information
– line / sample latitude / longitude
2. Conversion of geographic coordinates into map-projected coordinates
– latitude / longitude xmap / ymap
– choice of map projection and datum
3. Determination of a transformation function to map image coordinates into projection coordinates
4. Resampling using mapping function
– determination of pixel value in the map projected using one of the interpolation methods
Franz J Meyer, UAF
GEOS 657 Microwave RS - 31
Example: Original Image
Franz J Meyer, UAF
GEOS 657 Microwave RS - 32
Example: Transformed Image
(After Steps 1-3)
Franz J Meyer, UAF
GEOS 657 Microwave RS - 33
Example: Transformed Image
(After Steps 1-4)
Franz J Meyer, UAF
GEOS 657 Microwave RS - 34
Geometric Terrain Correction
• Geometric terrain correction (GTC) describes how to remove geometric distortions by using a DEM in the
geocoding process:
– To make sure that ALL pixels appear at their proper geographic location.
– To allow for overlaying SAR data onto remote-sensing data from different sensors
• GTC problem: What are the image gray values in every pixel of the output (geocoded) image given the input
image and the DEM?
• Two main approaches for geocoding including GTC are shown in the following:
– Backward Geocoding
– Forward Geocoding
Franz J Meyer, UAF
GEOS 657 Microwave RS - 35
Forward Geocoding
Forward geocoding
• Forward geocoding starts from a pixel in the slant-range
image for which the location on ground is determined.
• This results in an irregularly distributed point cloud on
ground and a subsequent resampling step is necessary.
• Terrain correction is done in image space.
Franz J Meyer, UAF
GEOS 657 Microwave RS - 36
Backward Geocoding
“painting the DEM in map-projection space”
• Starting from a predefined pixel location in the geocoded output image, the
corresponding column and line in the slant-range image are calculated using the DEM
information.
• The grey value for the output pixel is calculated through interpolation in the slant-range
image.
Franz J Meyer, UAF
GEOS 657 Microwave RS - 37
A Second Look At Forward Geocoding
1. Forward Geocoding:
• Step 1: Project pixels (xi,yi) into output image go using DEM
• Step 2: Determine gray values in output image by interpolation between
projected pixels
1. Project into output image 2. Interpolate gray value in
output image
Input image Output image
Franz J Meyer, UAF
GEOS 657 Microwave RS - 38
A Second Look At Backward Geocoding
2. Backward Geocoding:
• Step 1: Project output image pixels (xo,yo) into input image gi using DEM
• Step 2: Step 2: calculated gray value of projected pixel by interpolation
and assign to (xo,yo) in output image
1. Project into input image
3. Assign gray
value to pixel in
output image
2. Interpolate
gray value in
output image
Input image Output image
Only Inverse Mapping guarantees that ALL pixels in the output image receives a gray value (no holes) → inverse mapping is
standard approach
Franz J Meyer, UAF
GEOS 657 Microwave RS - 39
Geometric Terrain Correction Example (I)
Original Image
3
McKinley National
Park Airport
ALOS PALSAR (L-band) data © JAXA (2007)
Franz J Meyer, UAF
GEOS 657 Microwave RS - 40
Geometric Terrain Correction Example (II)
Geometrically Terrain Corrected Image
3
McKinley National
Park Airport
ALOS PALSAR (L-band) data © JAXA (2007)
Franz J Meyer, UAF
GEOS 657 Microwave RS - 41
Radiometric Terrain Correction (RTC)
• Problem: Sensor facing slopes appear overly bright in radar images.
• Cause: Pixel Size on sensor-facing slopes is larger → more ground is integrated into pixel →
brightness goes up
• Solution: Radiometric Terrain Correction (RTC)
1. Using DEM and observation geometry, calculate exact equivalent area 𝐴𝜎
covered by each pixel
2. Normalize radar cross section by 𝐴𝜎 to arrive at terrain normalized data 𝜎𝑇0
Franz J Meyer, UAF
GEOS 657 Microwave RS - 42
Radiometric Terrain Correction Example (I)
Image after GTC
3
McKinley National
Park Airport
ALOS PALSAR (L-band) data © JAXA (2007)
Franz J Meyer, UAF
GEOS 657 Microwave RS - 43
Radiometric Terrain Correction Example (II)
Image after RTC
3
McKinley National
Park Airport
ALOS PALSAR (L-band) data © JAXA (2007)
Franz J Meyer, UAF
GEOS 657 Microwave RS - 44
VV/VH/Ratio
VV/VH/Ratio
Google
Pixel Earth
RGB
RGB
scatteringImage
composite
composite
area of a 85before
after
x 85
calibration km 2
pixel
pixel
factors
derivedscattering
region in Western
from SRTM normalization
Arkansas
areadigital elevation data
Example of RTC Normalization for an Area in Arkansas, USA
Franz J Meyer, UAF
GEOS 657 Microwave RS - 45
RTC Example
Comparison of Processing Techniques
GTC Image
[not RTC processed; reference]
Franz J Meyer, UAF
GEOS 657 Microwave RS - 46
RTC Example
Comparison of Processing Techniques
RTC Image
[Processed with GAMMA RS]
Franz J Meyer, UAF
GEOS 657 Microwave RS - 47
Sentinel-1 RTC images over El Salvador and Honduras
El Salvador and Honduras
Franz J Meyer, UAF
GEOS 657 Microwave RS - 48
Sentinel-1 RTC images over El Salvador and Honduras
Puerto Lempira, Honduras
Franz J Meyer, UAF
GEOS 657 Microwave RS - 49
Environmental Monitoring using RTC Images
Example: Dynamics of Areas near Galena, AK
• Radar brightness times series of locations near Galena, AK
1 2
4
Below Freezing
River Water Tundra [high(er) moisture] 1
2
3 4 3
Below Freezing
Sentinel-1 SAR Time Series Yukon River, Alaska
Tundra [low(er) moisture] Lake Water
SAR Backscatter Time Series Information
Franz J Meyer, UAF
GEOS 657 Microwave RS - 50
What’s Next?
• Lab week:
– Labs on SAR image focusing and geocoding
– Labs on SAR image time series visualization and analysis
• After that, we will start talking about InSAR, a method for topographic mapping and deformation analysis
• In preparation for next week, please read the following information in the SAR Handbook:
– Chapters 2.1, 2.2, 2.5 & 2.6.1
Franz J Meyer, UAF
GEOS 657 Microwave RS - 51