Multilooking necessary when geocoding?

Hello,
I am fairly new to the SAR processing and I am trying to learn the basics through the ESA Sentinel Toolbox tutorials, which are a valuable resource.
In the very first tutorial “SAR Basic Tutorial”, I don’t understand the role of the Multilooking process specially when compared with Geocoding.
According to the tutorial, the “Multilooking can be an optional step, since it is not necessary when terrain correcting an image”. But in the Terrain correction step of the tutorial, the input image is the multilooked and speckled image.
So what I would like to know is if the Multilooking is actually a optional step or it is usually applied in pre-processing.
Thanks in advance
Niccolò

Multilooking is not mandatory. It fixes the inconsistent pixel sizes caused by the angle of the incoming signal and produces an image with “squared” pixels. As a nice side effect, speckle is reduced to a certain degree.
When you cange the number of looks in the module you see the changes in your total resolution. If your data has a good resolution, TSX for example, you can apply multi-looking to reduce it to about 10 meters and still have sufficient resolution for most applications. However, if your data resolution is already coarse or if you need the spatial resolution for the search for certain objects I would personally advise to skip the multilooking.
As you said, the final geocoding is performed in the RangeDoppler Terrain Correction module at the end of your pre-processing. You can apply it on both multilooked or non multilooked data and (besides of some resampling) it won’t change your resolution any more.

10 Likes

Thank you for your quick and exaustive reply!

But if the data is GRD, the pixel is square. It is not like this? Correct me if I’m wrong. Thanks

yes, after multi-looking the pixels are of equal spacing in range and azimuth direction.

Multi-looking has dramatically reduced the duration of Snaphu export from nearly 15 minutes (without applying multi-looking to the interferogram created from S1 SLC IW data) to 5 seconds (after applying multi-looking). Does that imply that multi-looking may cause unwrapping errors?

You can only confirm that by looking at the interferogram before and after unwrapping. In an ideal case, multi-looking reduces the file size of the input image while maintaining the fringe patterns which are to be unwrapped. If these patterns are very fine, multi-looking can lead to loss of this information, depending on the pixel size and the spatial scale of the patterns.
In turn, applying multi-looking on phase noise neither increases nor decreases the unwrapping quality (garbage in > garbage out).

So the processing time alone won’t tell you much about the success or failure of the process, you have to check if what has been unwrapped makes sense.

Is it possible if you could give me an example of what you mean by fine fringes pattern ?

Didn’t we have this discussion already?
Edit - here: Could you suggest me a substitute for STAMPS software for performing Persistent Scatterer interferometry (PSI)? - #4 by ABraun

Oh we did. I am sorry, my mistake. I will make sure to research more before asking :slightly_smiling_face:

No worries - more examples here: Phase unwrapping and low coherence - #2 by ABraun