Histogram changes during interferometric processing chain

Microwave Toolbox Interferometry
#1

Dear All,
Considering an interferogram, after merging, over an IW Sentinel-1 images couple (S1A_IW_SLC__1SDV_20160909T050422_20160909T050450_012971_014864_149C and S1A_IW_SLC__1SDV_20160921T050423_20160921T050450_013146_014E2A_06C0) the following processing steps have been applied:

  1. Topographic phase removal (see attached topoPhaseRemoval_test.xml (1.4 KB) )
  2. Multilook (see attached multilook_test.xml (1.3 KB) )
  3. Terrain correction (see attached terrainCorrection_test.xml (3.0 KB) )

and the accurate histogram of the interferometric phase has been computed after each step.

It can be noted that the phase’s histogram is initially flat:

3burst_Split_Orb_Back_ESD_Ifg_Deb_Merge_hist.PNG1324×619 35.4 KB

After the topographic phase removal the phase distribution starts to be no longer flat:

3burst_Split_Orb_Back_ESD_Ifg_Deb_Merge_DInSAR_hist_2.PNG1877×927 48.2 KB

This is the histogram after the Multilook:

3burst_Split_Orb_Back_ESD_Ifg_Deb_Merge_DInSAR_ML_hist.PNG1877×932 50 KB

The histogram after the Terrain correction shows an increased distortion:

3burst_Split_Orb_Back_ESD_Ifg_Deb_Merge_DInSAR_ML_TC_hist.PNG1876×925 50.9 KB

Are these changes of the phase distrIbution through the processing steps expected?

#2

Make sure the Multilook is not multlooking only the phase band. Without any bands selected, it will multilook i and q and produce a new virtual phase band.

#3

Hi Luis,
Thank you for the quick answer.
I can confirm you that the multilooking is performed without selecting any source band.
However a phase distribution distortion can be noted just after the TopoPhaseRemoval, so the distortion starts before the multilooking.

I’m available to provide you any other info that can be useful for this analysis.

Thank you again very much.

#4

Luis,

this problem exists and we are considering it of high importance because in the procedures following the interferogram production, the phase “rotation” could not be trusted. Please have a look at the histograms of an example. I confirm that the multilooking is performed without selecting any source band, following your advice in a past post.

thank you

Panagiotis Elias

problem_SNAP_non_physical_phase_histogram.png960×720 50.8 KB

#5

I do not see any issue. There is no probelm in having ifgs with non flat histo (wrapped phase values).
In fact, in that specific case it seems that the topo phase removal did correctly its job by showing up information hidden below the noise level of the ifg. In addition, the multi-looking improoved this info further by seperating it more from background noise. Finally, the histo of the geocoded product looks very similar so no issues there as well.
You should make sure (also mentioned before) to multilook i&q and not the phase directly as well as when TC to select the option for complex products.