I don’t know if this causes the problem but filtering before coregistering doesn’t seem good to my. How can you perform step 4 MultiTemporal filtering without coregistering?
You are right in fact I have read the help file of MultiTemp filter which says “For a sequence of N registered multitemporal images, with intensity at position (x, y) in the kth image denoted by Ik(x, y), the temporal filtered images are given by…”
Therefore, the images should be first co-registered.
I have worked a lot on single image analysis with the steps (without subset and coregistration) but I need to test my model (which is not related to interferometry or polarimetry) for time series data sets as well.
as coregistration automatically searches for significant pairs of points in the images, I would personally leave out the filtering and apply a real multi-temporal filter on the stack after coregistration. Would be worth a try.
But what about the subset phase. I generally subset the images after Range Doppler Terrain Correction but now I have to process about 100 images soo what do you think, can I do the subset before Radiometric terrain flattening in order to save some more time?