Hi I want to upsample s2 bands from 10 – 60m to 10m resolution. Which processor is recommended S2resample or Superresolution?
The S2processor takes less time but the output images look noisy. However, after applying C2RCC there is a very small difference in the Reflectance spectrum and Chl-a concentration between the two versions (60m and 10m) resolution.
can you give an example on what you mean by noisy?
The upsampling of the S2 Resampling operator simply uses local image statistics to make the resolution higher while superrres uses the variation from the 10m bands to kind of burn it into the 20 m and 60 m bands.
As you later on apply the C2RCC operator, you cannot really expect to get a fully reasonably 10 m resolution of the output, as several bands are involved.
S2A_MSIL1C_20180410T112111_N0206_R037_T32VMR_20180410T132753_s2resampled10_B2.bmp (1.3 MB)
S2A_MSIL1C_20180410T112111_N0206_R037_T32VMR_20180410T132753_s2resampled60_B2|690x383,50%
Please see the attached two images, 60m resolution and 10 m resolution.
thank you.
I don’t quite understand: Band 2 is acquired at 10m, why is it resampled to 60m?
Sorry, the attached two bands are processed through the S2Reample processor in SNAP. 1) Sentinel 2 Bands with Band2 (spatial resolution of 10m) and 2) Band2 Upsampled to 10 m resolution. The bands are resampled to 10 m as C2RCC requires all the bands to have the same spatial resolution. Band 2 has already a spatial resolution of 10m, so it won’t be upsampled but if you see the attached images, the resampled version looks noisy.
S2A_MSIL1C_20180410T112111_N0206_R037_T32VMR_20180410T132753_B2.bmp (1.3 MB) S2A_MSIL1C_20180410T112111_N0206_R037_T32VMR_20180410T132753_s2resampled10_B2.bmp (1.3 MB)
I think is normal that the 10m resolution band is noisier, because of the finer resolution. When you resample to 60m there is some statistical procedures that “averages” the noise. But your scene is quite strange, like a border of a swath or something. Did you subset your area?
No, I didn’t subset my data. I am using the whole scene because my insitu lat/log is changing for every scene.