I need to conduct a Landuse classification of New Caledonia. I therefore used the Landsat 8 satelite images of one year and created a multi temporal mosaic by a mean function. As most of the areas were ostly clouded, I used a cloud mask, which resulted in areas of no value. I then tried the focal.median function. However, this led to a change of DN values for the whole image.
I am therefore looking for a function which only tackles the specific areas of no data and leaves the other values untouched.
I was thinking of using Sentinel-1 Radar data, check the radar signature of those areas of my landsat8 image where I have no values, look for the same radar signature at another area, from where I also have a spectral value in my Landsat8 image to then use that value of my Landsat 8 image to fill my missing Landsat8 areas.
Does that make sense? If so, how could I execute that idea? Or is there even any other option?