I am currently working on surface subsidence using a stack of SENTINEL-1 images. I understand that the more images we stack, the more the atmospheric error is minimized.
So, my question is:
Lets say i have five images, perform the interferometric processing (pair1-pair2, pair2-pair3, pair3-pair4,pair4-pair5) and create a stack out of them. pair1-pair2 form the
subsidence map1, pair2-pair3 form the
subsidence map2, pair3-pair4 form the
subsidence map3 and pair4-pair5 form the
subsidence map4. Finally, i add them all up (
subsidence map1 +
subsidence map2 +
subsidence map3 +
subsidence map4) to find the final displacement map. At the very end, i want to discard pixels values that have coherence value below 0.5.
So, do we just choose a random coherence layer out of the 5 pairs to masks out the low coherence values?.
Do we add up the coherence layers as we did for the subsidence map?
Do we multiply each coherence layer and the we mask it out?
What method is recommended to follow?