Dear friends,
Now, I use SNAP 5 to process sentinel GRDM data. I want to do speckle filtering using batch processing, but in the graphs of batch processing, I could not find the speckle filter.
Could you please show me the way? or there is no batch for speckle filter?
Kind regards,
Liyun
Hi @dlydai,
First thing before doing batch processing is to create a graph with the operators you will use later.
First go to Tools -> Graph Builder, create your graph and SAVE it. You can see on the image where the speckle filter operator is located.
Once you have saved the graph, you can close it. Open now the batch processing tool (Tools -> Batch processing) and open you graph. You will see now that the operators of your graph are showing. Last thing, load the data you want to process, check your output settings and press Run
If you have a multi-temporal series of SAR images you could also consider Multi-temporal speckle filter (for this you need to create stack with the time series first)
Having a time series of coherence images obtained from S1A with the same orbit (same acquisition geometry then), would it be enough to create stack or would it be more precise to run coregistration before multi-temporal speckle filtering?
M
if the intensity is still int he product, coregistration is the better choice. If you only have coherence, it probably won’t work, so you need to terrain correct and create a stack instead.
Why co-registration is better if intensity is still in the product? What makes the difference? In my case I only have coherence so steps would be then: TC each coherence -> stack -> multi-temporal spk?
Isn’t better to apply spk filter before TC? What about stack coherence images -> multi-temporal spk -> TC
So far, I have tried this without any error message:
M
coregistration searches similar patterns in two images to find their best alignment. Coherence has nearly any patterns which occur in both images so this will technically fail.
Speckle filtering your coherence also makes little sense to me because speckle filters are designed to be applied at the backscatter intensity. If you have many images, multi-temporal makes sense at the intensity level. But it makes nearly no difference for less than 6 images if you apply a single or multi-temporal filter.
See here: Single or Multi-temporal speckle filter?
Multi-temporal filtering of coherence-stack should work even though it’s stretching the assumptions of the Quegan filter. One should choose a generic filter like median for the spatial filter.
yes, nothing wrong with smoothing coherence in the first place.
Great, thank you. My target are urban areas so I guess spk filtering is not that important since values are pretty high compared to other classes. Makes more sense for me TC and stack but I’ll check if there is any notable difference by adding the spk filtering step.
M
