How to stack deramped SAR images in SNAP 7.0?

Dear SNAP developpers,

Thanks you very much for the new version of SNAP, and especially the deramp operator.

From what I can see, the operator on splitted TOPSAR SLC images works quite well (left = before deramp, right = after) :

However, when comes the time to coregister the images, problems arise.

It is clear from the help that the stack of TOPS SLC images cannot be deramped: “The input to the operator cannot be deburst product or a stack of products such as the output of Back Geocoding operator.”

It means that a process of this type is not allowed :

My idea was then to deramp the images just after the split (as suggested by the help) and then back geocode the image. Like this

The process worked like a charm, in appearance, but results show an azimuthal discrepency in the slave image

In addition, computing an interferogram on a “classical” SLC stack or on a deramped SLC stack gives different results.

I’m not sure if it is intentional or not, but working on deramped TOPS SLC images alone serve no purpose in my knowledge. However, I see no way to deramp a stack of TOPS SLC images or, equivalently, stacking deramped TOPS SLC images.

Thanks for reading,

Quentin

1 Like

@jun_lu & @lveci should be able to help.

If your goal is to generate interferogram, then you don’t need to perform deramp/demodulation because Backgeocoding has already done it for you. Inside Backgeocoding, deramp/demodulation is performed on the slave image before interpolation, reramp/remodulation is performed after the interpolation. You got artifacts in the output image from your second graph because you have removed the phase ramp and modulation and Backgeocoding was trying to do it again.

The deramp/demod operator is for users who have not intention to generate interferogram and just want to generate a debursted SLC image. In this case, if no deramp/demodulation is performed, artifacts will be introduced in the later on processing such as interpolation.

1 Like

Dear @jun_lu,

Thanks a lot for your response, but it quite does not answer my question.

I totally agree with you about that, but my goal is not to perform interferometry.

Since deramping only affects the phase, I honestly don’t see any application where only the phase of one single SAR image is used.

May I re-ask you the question : How to stack deramped SAR images in SNAP 7.0?

Thanks in advances ,

Quentin

Dear Quentin,
I probably misunderstood your question. Regarding your last question on creating coregistered (I assume by stack you mean coregister) deramped images, I don’t think currently you can do it in SNAP 7.0. This requires modification on the Backgeocoding operator, basically (1) adding a step to deramp/demod master band and (2) skipping the step of reramp/remod slave band. Then the output of Backgeocoding is deramp/demod stack. The other way to fix this is that we can modify the deram/demod operator to undo the reramp/remod applied to slave in Backgeocoding on condition that all information needed are available in the metadata. We can discuss this issue in our next meeting with ESA to see when we can make the modifications.

Sorry about that,
Jun

1 Like

Dear @jun_lu,

Thank you very much and no need to be sorry.

Yes, I should have been more precise.

I have a parallel SNAP version built in IntelliJ-IDE where I tried to modify the code to perform what you’re saying. Unfortunately the way it is written does not give much flexibility to do that (without changing a lot of lines). I will try again and let you know if it works.

Cheers,

Quentin

Hi @jun_lu,

I’m coming back to you after some new results. I “managed” to coregister the deramped TOPS images.

I debursted the deramped images so that I can perform more common coregistration. Coregistration gave me a java_null_pointer_exception. Then I tried the DEM-Assisted Coregistration. It worked quite well. To be sure the phase information is correct, I performed interferograms using my stack of deramped TOPS SLC images. The results are pretty I guess.

I compared this interferogram with the one computed using the “classical” Sentinel-1 chain in SNAP (S1 TOPS Coregistration), meaning on reramped TOPS SLC images.

When I look at the difference between interferograms (mod 2 \pi), there’s for each burst a residual phase pattern. This phase pattern can be important.

1 Like

What is your use-case exacly Quentin?

Marcus

I’m sending you a private message.

A ticket has been created to capture the issue: https://senbox.atlassian.net/browse/SITBX-640

1 Like

Jun, can we extend the use-case so that deramping co-registered stacks becomes possible?

Yes, we can modify the Backgeocoding operator to make it output deramp/demod stacks

1 Like

Does this means that classical Insar processing chain in SNAP introduces linear phase ramp in random direction for every burst, or could there be another explanation?

I wish I knew. Because my area is moving fast in azimuthal direction, other biases related to the TOPSAR acquisition mode does not allow me to isolate the problem.

I believe the ramp phase has been cancelled out during the calculation of the interferogram

Just a notification to say that it now works like a charm on SNAP 7.0 public release. Thanks for the hard work

EDIT : SORRY BUT STILL NOT WORKING. Still azimuthal phase bias :

@qglaude I follow this topic from the beginning and would be interested in what you are trying to do with these deramped stacks. So feel free to keep us updated in here about the outcomes :slight_smile:
(of course it is understandable if you want to keep it secret until you have officially published something with it)

I will :slight_smile:

Right to the point. I’d like to be transparent but not possible until stuffs are at least submitted :confused:

Don’t feel pushed - as it is your work you have to make sure that you are the main benefitor in the first place. You can also simply post a link to the publication once it is available.

The Backgeocoding operator has been modified to add an option to output deram/demod master/slave bands. You can give it a try.

1 Like