Enhanced Spectral Diversity Operator

I think I found a problem (and a possible solution) with the ESD operator.

As I have already shown in this forum (see here) I have been getting strange artifacts correlated with the bursts’ borders in my interferograms.

Here the coherece and interferometrc phase (S1 TOPS Coregistration -> Interferogram Formation -> S1 TOPS Deburst -> Multilooking):

By searching on the forum for similar issues I found a topic where was presented a similar problem (see here). A interferogram with strange artifacts related with the bursts’ borders. I asked the name of the used products to repeat the processing and I got a similar strange result:

From the literature I learned that ESD should overcome this problem so I changed my processing chain by adding the ESD operator after de S1 TOPS Coregistration and I repeated the processing for both data sets and I got the following results:

These results present a clear degradation of coherence.
I noticed that these new strange patterns in coherence were correlated with the deramp demod phase. Therefore I decided to study the source code in order to find out why this was happening. In my study I noticed the ESD operator was using the deramp computed in the previous operator (back-geocoding). So I changed that by computing a new deramp demod phase before the ESD operator perform azimuth shift.

When applied to the data I got the following results:

As you can see I solved the problem with the phase decorrelation caused by ESD operator.
And the azimuth borders pattern is solved for the second data set.


Great thanks. @junlu is currently on vacation. He’ll look into this as soon as he is back.

Hi all,

I have found similar problems with the burst border edges using s1tbx. I coregistered two (merged) images, processed by IPF 002.71:


20160729 (master):

Although the phase information looks good, the coherence shows decorrelation at the burst edges:

Raw phase (no topo removal)


Are these issues also related to the deramping function?

Thanks in advance!!


Have you used the ESD operator after the coregistration?

Yes we used the ESD operator after back-geocoding.

Then I think your problem is the same I have described above.

I think they are working on this issue, for they asked me more info about it.

@lveci, Have you any news on this?