Yes, using two polarisations from the same acquisition is not multi-temporal so I would assume that the filtering is less efficient , and in theory perhaps even incorrect (due to differing polarisation). But in practise I would guess that you can get away with it (and need to use more scenes).
The multitemporal filter assumes that the features in the scene are stable. If you have a stack over tidal flats for example with the shoreline being in a different position is each acquisition, the filter will smear the shoreline in all images.
Okay. So if the land cover I am looking at seems to be stable, as many as possible scenes could be used for filtering. If changes could occur, e.g. small scale deforestation, storm damage or a different land use (crop rotation), I shouldn’t use multitemporal speckle reduction and filter each image individually? What about using 7 additional images for multitemporal filtering (operational use by temporal window) and going back to single speckle filtering, just in case something strange appears in the datasets after filtering? I want to have the opportunity doing individual scene analysis, based on the multitemporal filtered result.
Just inspect your stack to see i there are large changes and adjust your processing if necessary. - in most cases the multitemporal filtering should be the valid approach. If there is a large sudden change you can split the processing in pre-event and post-event stacks.
Thanks for your ideas!
I did multitemporal filtering using 20 S1 datasets in total (10 acquisition dates). Included were asc & desc scenes (TF applied). I found a position with a huge change, but the result looks good. What do you think about it? Obviously it doesn’t make so much difference?
Mixing ascending and descending scenes in multitemporal filtering is not advised since they will have different ground backscatter and geometric shift of objects due to the different view points. Mixing co- and cross-polarized images is also not advised.
I have made a simple filter and a multi-temporal filter for ALOS PALSAR (7 images) and S1B (11 images). I would like to know how I can compare the speckle reduction in both filters. How is the ENL calculated? On which areas should I calculate it? The images I used are for small areas of the Peruvian Amazon.