i’m trying to fuse Sentinel1 GRD data and Landsat8 data. So I performed an atmospheric correction on the multispectral data to achieve reflectance values. The workflow for the Sentinel data consists of the calibration to sigma nought, speckle reduction and range-doppler terrain correction. But these steps raised some questions:
If I would fuse these datasets, do I have to apply the “linear to dB”conversion? Or should I use the values resulting from the range-doppler terrain correction? Or do I have to rescale the values according to the bit depth of the multispectral data?
If you want to combine them visually doing the linear-to-dB is a good idea (a linear SAR image has much more dynamic range than can be seen with the human eye).
The aim of the fusing process is not only to compare them visually but to integrate the sar data into the multispectral dataset to form a new dataset which contains the information from both input data sets.
Linear-to-dB conversion makes the histogram much more similar to your optical channels so I’d say it makes great sense to do that transformation. It’s reversible in any case.
Hi walter
I am preparing to combining S1 and Landsat-8 and following your steps. I have some questions. Should I co-register the S1 terrain corrected image and Landsat-8 image first? If co-registering is not need, the accuracy between two images could be enough? If you performed the co-registering, how did you do that?
Thank you a lot!