Processing - ASAR

1. SAR Software

There are several software packages available for image processing of raw Synthetic Aperture Radar (SAR) data, the three evaluated for ASAR image processing were: -

• Gamma Software to support the entire processing of SAR data (SAR processing, InSAR, DInSAR, classification), supports ASAR Image Mode and Alternating Polarisation Mode.
• Roi_Pac Repeat Orbit Interferometry Package from Caltech/Jet Propulsion Laboratory, used mainly for interferometric SAR processing, support for ASAR Image Mode only.
• SARScape SARMaps specialized ENVI/ArcView extension for processing of SAR/InSAR data, supports all three modes of ASAR data.

For purposes required at the Landmap Service the SARScape software was selected for the image processing. SARScape software users either ENVI or ArcView user interface. The ENVI implementation of SARScape was used for this work.

2. ASAR Data Selection

The ASAR scenes have been selected for British Isles coverage using ESA EoliSA client. This can be downloaded free of charge and imagery can be viewed using an anonymous user log in. If you require a particular ASAR image which is not available from the Landmap Service then please send your requests to This e-mail address is being protected from spambots. You need JavaScript enabled to view it including the data's date & time of acquisition, mode, co-ordinates, size and coverage. Use the Save set function of EoliSA software and we will try and get the data you require.

3. ASAR Processing

3.1 Processing ASAR from Raw

Initially when ASAR data is received from the European Space Agency (ESA) the data is in a raw format on a DVD and is a level 0 product (0P). The DVD contains information about the start date and time and stop date and time the product was acquired along with the orbit identification number. The product type is also specified (Table 3.1).

Table 3.1. ASAR Product code descriptions as used by ESA

The processing steps below are described in more detail within the help manual of SARscape software. The processes described are specific to those perform at the Landmap Service to produce the GeoTiff's available for download.

Step 1 ASAR Import & Focusing

A level 1 focused image was generated from the ASAR RAW data.

SARscape provides a drop down menu to select the product type (ASAR WS, ASAR AP & ASAR IM) in order to import Raw Radar Data. Supporting files are required for this process, these are: -

The orbit file (DOR_VOR or DOR_POR) measured by the Doppler Orbitography and Radiopositioning Integrated by Satellite (DORIS) instrument for the day of the input dataset. ESA should be contacted to obtain access to these files.

ASAR XCA, INS and XCH (WS only) auxiliary files, provided by ESA from, http://envisat.esa.int/services/auxiliary_data/asar/, these are also provided with SARScape. The file corresponding with the date of the image acquisition was used.

The data file from the DVD was then imported.

The output is a Single Look Complex (SLC) image.

Step 2 Focusing and Multilooking

The input file is the slc file created in step 1. Focusing was performed in the previous step so the multilooking option was selected by checking the tick box. The Looks button was clicked so that SARscape could calculate the appropriate multilooking factors to apply to the processed data. Then an output name was given, appending _pwr extension at the end of the process to produce an intensity image.

Step 3 Conventional Despeckling

Noise contained in Synthetic Aperture Radar (SAR) data can be reduced by using a spatial filter for a single image. Gibson et al., 2000, p.161 defines filters as 'a mathematical procedure that alters the digital numbers in an image'. Spatial filters are used to emphasize or deemphasize image data of various spatial frequencies (Lillesand et al., 2004). There are generic filtering algorithms available which can be applied to both optical and SAR data e.g. Mean, Median and Mode. There are also SAR specific filtering algorithms one of which was select for the image processing.

The Frost filter was applied to the ASAR data that is provided on-line as 8 bit and 32 bit greyscale GeoTiffs. The SARscape manual described Frost filter as, 'a non-linear filter which reduces speckle noise as a function of the degree of heterogeneity measured by the local coefficient of variation calculated within the window'. Other options provided are Lee and Refined Lee. The rest of the settings were left at default. The intensity image _pwr was used as the input and the output filtered image was produced with _fil extension.

Step 4 Geocoding and Radiometric Calibration

The geocoding process transforms the filtered image which is in image coordinates to geographic coordinates in this case to British National Grid. A range-Doppler approach was used to geocode the ASAR images. No Ground Control Point (GCP) was selected however a sub-pixel accuracy can still be achieved when DORIS data has been used as part of the processing.

75m SRTM data provided the elevation data needed for the geocoding process and an output grid size of 25m was defined. A bilinear resampling method was used and radiometric calibration option was selected. The output is a geocoded image _geo which can be viewed in ENVI 4.1

Step 5 Produce 8 bit & 32 bit Greyscale GeoTiff

The geocoded file (extension _geo) was viewed in ENVI 4.1. SARScape was used to create the 8bit GeoTiff images. Envi was used to output a 32 bit GeoTiff images.

3.2 Pre-processing Steps for ASAR Interferometric Pairs

Interferometric pairs (images of the same geographic location) were ordered from ESA and two intensity images (_pwr) were created by working through Steps 1 & 2 in section 3.1.

Step 3 Amplitude Coregistration

An interferometric pair was used to create an RGB colour composite image therefore coregistration is required as in later processing steps a time series filter will be applied to the images for despeckling and then an RGB composite created.

One image from the pair is used as the input reference file i.e. a file to which the other input file is coregistered to. Then the second image file was selected for the input file option. The bilinear resampling method was used. The output of this process is an image with _rsp extension which indicates that the images have been resampled and resized and coregistered with the reference image.

Step 4 Multi-temporal Despeckling

For a multitemporal data set a time series filter is recommended to reduce speckle. The coregistered interferometric pairs provide the input files with the output files given the extension _fil. The Equivalent Looks variable was set at default -1, this means that the Coefficient of Variation (CoV), which determines this thereshold for the speckle filtering, is calculated by the software. The CoV is defined, for amplitude images, as 0.5227/sqrt (number of looks).

Step 5 Geocoding and Radiometric Calibration

The same process was carried out as that described in step 4 section 3.1 for the interferometric pairs. The output being a pair of images with _geo extension, this is the correct format for generating RGB colour composites (see section 3.4).

3.3. Creating a Coherence Image

Step 1 in section 3.1 was performed to create a single look complex image for the interferometric pairs so that the images are in the correct format for extracting coherence.

Step 2 Feature Extraction

A coherence image extension _cc was created by calculating the degree of coherence between the two slc images produced in step1. For a coherence image to be produced the two slc images need to be suitable for interferometric processing.

Step 3 Geocoding and Radiometric Calibration

The same process was carried out as that described in step 4 section 3.1 for the coherence image producing a coherence image file with _geo extension which can then be used to generate unsigned and signed coherence colour composites (see section 3.4 methods B & C).

3.4 Generating RGB Colour Composites

For RGB Colour Composite Processing coregistered or geocoded input products are mandatory.

A) Multi Difference Colour Composite

Requires the use of two images of the same geographic extent but aquired on different dates which have undergone amplitude coregistration.

input 1 = amplitude image 1 (usually the older dated image of an image pair)
input 2 = amplitude image 2

Bands Assigned: -

Red = input 1 - input 2
Green = input 2
Blue = input 1

B) Unsigned Coherence Colour Composite

Requires the use of three images of the same geographic extent one of which is the coherence image and two amplitude images.

input 1 = coherence image (measure of change between the two dates)
input 2 = amplitude image 1
input 3 = amplitude image 2

Bands Assigned: -

Red = input 1
Green = (input 2 + input 3) / 2 (mean amplitude values)
Blue = input 2 - input 3 (difference amplitude values)

C) Signed Coherence Colour Composite

The process is the same as creating an unsigned colour composite image apart from the images used for the inputs are in signed rather than unsigned format.

4. Overview of ASAR Processing Chain

Product 1 - Grayscale ASAR Images

Product 2 - Multi-Difference Colour Composite Images