Has anyone had any experience or success running this package SAR2SAR to despeckle Sentinel-1 images? The repository provides the pre-trained weights to run locally but I have not had seen any success in the outputs yet and was wondering if other had any experience.
Thank you,
Andrew
This project appears to be an attempt to repackage a well-established technology as something innovative by using buzzwords.
We can utilize the persistent scatterers function or ADI index in conjunction with a weighted Gaussian filter to effectively remove speckle noise. This approach assigns higher weights to temporally stable pixels. For a detailed explanation and a practical implementation using PyGMTSAR (Python InSAR), you can refer to this video lesson: https://m.youtube.com/watch?v=rIbdVeu8qO0
Thanks for the response! Is PyGMTSAR still applicable if I am not doing InSAR? My goal here is to get an amplitude image of Sentinel-1 while minimizing any filtering which is why I thought this SAR2SAR package could be beneficial.
Yes, certainly. I have implemented a set of scalable algorithms in PyGMTSAR to efficiently process radar data on a wide range of hardware, from 4GB RAM Docker containers to workstations with 512+ GB of RAM. The stable version of PyGMTSAR is capable of SBAS processing and trend analysis. Additionally, the new development version aims to expand its capabilities to cover amplitude image processing, PSI, SBAS analysis, and more. You can find the project on GitHub: GitHub - mobigroup/gmtsar: PyGMTSAR. If you explain your desired output, I can provide you with an example.