QuadPol decomposition procedure? (When to filter?)

I’ve been trying to decompose Quadpol data with SNAP and it kind of works.

The processing chain as of now looks like this:

Calibrate -> Multilook -> Matrix_T3_generation -> Decompose (Yamaguchi) -> Geocode (with creating a mosaic) -> Export tiff (then opening in other software for sampling)

The problem is, I have to filter somehow. In PolsarPro I apply Lee Refined filter after I create the T3 matrix. But in SNAP I get an “unexpected error”.

I tried to filter before creating the matrix but the data looks terrible after it.
I can’t do all of this in PolsarPro though since I can’t geocode there.

Has anyone done this and what is the correct or optimal ordering of processes?

What I need to do in the end is to extract values from the 3 types of decompositions from sample sites outlined by polygons.
I am doing this step in Arcmap because I already have a good model for extracting data. But Maybe I can work around it if extracting data can be done in SNAP?

With geocode you mean performing a terrain correction, right?

Not really, because SNAP simply can’t terrain correct SLC quadpol data downloaded from ESA (only from ASF)
older question I posted regarding that.
The thing it can do though is to write the geocoding by a workaround, that is mosaicing only one image. In my case it is ok since it is a really flat area and my specific areas are actually exactly where then need to be without any foreshadowing and such.

So in my case geocoding is not terrain correcting. But that shouldn’t really matter.
Even if we remove the geocoding or terrain correcting step, the issue is still that I need to speckle filter so I can analyze the results.

Ah ok, thank you for clarification. Sorry, I don’t have anything to help you with this issue…

After decomposing quad-pol RADARSAT-2 data, the values of all the components are in db scale (like -15 db). Can I just convert it to linear scale using db to linear? Will this give me the same output which I get from PolSARpro ?

I think you used the wrong filter. It should be the Refined Lee Filter under Polarimetric Speckle Filter. Also, you probably want to do multilook after the decomposition.

You can convert the band in dB to a new band in linear scale by right clicking on the band and select “Linear to/from dB” option.