When to apply speckle filter for S1A imagery

Microwave Toolbox Polarimetry
1

Hi there, I am pretty new to this forum and to SNAP and to radar imagery processing, and I cannot find a concrete answer to this after considerable searching and reading online (and on this forum before signing up), so would like to seek help from this platform.

I have been dealing with some S1A imagery lately (18 scenes spanning from 2021 to 2023, therefore no thermal noise removal has been performed), and after performing band math using the sigma0_VV and sigma0_VH bands, I am now facing an issue on whether to apply speckle filtering now, or should I redo the whole thing, with speckle filtering applied before band maths (or before range doppler terrain correction?)?

The preprocessing (in batch) on the SLC product shown below has been performed, which should be pretty standard as I have taken reference from this forum as well as from other sources:

  1. SNAP 12: Read → apply orbit file → TOPSAR split → calibration (export as sigma0 band) → TOPSAR deburst → TOPSAR merge (merge the sigma0 bands for VV and VH) → multilook → terrain correction (SRTM 1Sec Grid) → subset to ‘rectangular’ extent of AOI → Write (as NetCDF-CF)
  2. QGIS: further clipping for the ‘real’ AOI and stacking, as for some reason, SNAP was not able to do this specific clipping for me
  3. SNAP 12: Read → Band Maths for calculating some indices with the sigma0 bands → merge → Write
    → An example of index calculated with the SNAP band maths is RVI, calculated with 4*sigma0_VH / (sigma0_VH + sigma0_VV)

At the moment, I plan to compare the performance of several speckle filters, such as Refined Lee and perhaps also Frost in the AOI. But to be honest I have no idea if I have performed everything correctly. The only two things that I am sure are that 1) I prefer not to apply land-sea mask as the AOI is related to coastal areas, and 2) no conversion to dB should be performed before band math (although I am not sure if the indices should be converted to dB for visualisation).

Any help would be greatly appreciated. Thank you so much for your help in advance.

2

You did not mention your application. Using a speckle filter and parameterizing it depends also on the degree of detail, you want to keep.

3

Hi @radarlove
Thank you for your message. The main purpose for me is to calculate the relevant indices using sigma0_VV and sigma0_VH in S1A imagery, visualise them and have some analysis, on SNAP and/or other software.

As far as I am aware of, speckle filters help smoothing out the noise, and some also help better distinguishing the boundaries, which would be crucial for my analysis. I have heard that applying filters prior to calculating may impeded the accuracy of results, but at the same time, some information suggests that applying some sorts of filter can remove unwanted noise and enhance the desired signals for further analysis. So I am really confused about which to do first.

Thank you again for your help.