Estimating the appropriate number of looks when multilooking images in SARscape

This Help Article explains how a user can estimate the appropriate number of looks when multilooking RADAR images using SARscape.
 
During the multilooking step of the SARscape basic module, an SLC (Singe Look Complex) radar image is divided in different looks characterized by different frequencies. The different looks are then averaged.

There are two reasons of completing this process:

1. It will reduce the speckle due to constructive and destructive interferences between the different backscattered signals from the different ground targets. These interferences result in a grainy appearance of the image especially above homogeneous areas. The speckle is directly related to the SAR system acquisition.

2. It will allow you to obtain a square pixel on the output image: the azimuth resolution of a SAR image is usually better than the range one. The resulting pixels on SLC radar images are then not squared and the SLC image will usually have more pixels in azimuth than in range for a same extent in range and azimuth.



Why multilooking will reduce speckle and output a square pixel image:

1. Speckle reduction: one way to reduce the speckle on one SLC image is to increase its SNR (Signal to Noise ratio). Averaging different looks of one SLC image will keep the mean of the data and will decrease their standard deviation. The SNR corresponds to the ratio between the mean and the standard deviation of the data. If the standard deviation decreases the SNR will increase.

2. Square pixel computation: averaging the different looks of one single image will result in a loss of resolution. The higher the number of looks the lower the resolution. Applying the multilooking only in the azimuth direction will only reduce the azimuth resolution. The appropriate number of looks can then be evaluated in order to obtain an output azimuth resolution similar to the range one (see below).



How to determine the number of looks to compute a square pixel:


1. Manual computation: here are the steps to manually evaluate the number of looks to obtain an output image with a square pixel.


A. Parameter collections. The computation requires:

      * the incidence angle of the image: α
      * the range resolution of the image: Δr
      * the azimuth resolution of the image: Δx
These parameters can be collected from the SLC image header file in SARscape which has a .sml extension.

Example:  the requested parameters are saved in a file called filename_slc.sml (with filename the name of the file which contains the SLC image). The 3 requested parameters are the following:

      * IncidenceAngle = 23.1281316063522°
      * PixelSpacingRg = 7.80397367094829 m
      * PixelSpacingAz = 4.21068474688921 m


B. Computing the ground range resolution: see figure below which represents the range geometry vs the ground geometry

   

   Δrg = Δr / sin( α )

Using the parameters of the previous example the ground range resolution is 19 meters


C. Computing the ratio between the ground range resolution and the azimuth resolution to obtain the appropriate number of looks:


   Nb_look = Δrg / Δx

For the previous example the number of looks would be 5 in azimuth, and 1 in range. It will result of a ground range resolution around 19 m and an azimuth resolution of 21 m

D. Enter the number of looks manually in the multilooking interface of SARscape.



2. Automatic calculation using the Looks button of the multilooking interface in SARscape:

This button will estimate the number of looks for you. However it should be used carefully. Indeed the calculation you will obtain will strongly depend on the default parameters that are defined in the SARscape Default Values (SARscape > Default Values) and especially the Cartographic Grid Size defined in the GENERAL parameters. This value should be set to the ground range resolution.
Note that manually calculating the number of looks sometimes gives different results than using the looks button. It is due to the calculation that is done differently. The different results are both correct in fact. With our example the looks button propose 4 looks (instead of 5 manually) in azimuth and 1 in range. The user can then choose between a pixel very close to a square one with 5 looks in azimuth or keep a better azimuth resolution with 4 looks

   

___________________________________________

Reviewed by BC on 09/17/2014