C2RCC after Super Resolution Plugin

Optical Toolbox
#1

After using the S2 super-resolution plugin, the C2RCC giving an error message that the source must be S2 MSI L1C product. Can we use c2RCC after super-resolution?

Asim

#2

I think C2RCC only works with resampled L1C data.

Besides that: The super-resolution algorithm is mostly just a visual improvement of the lower-resolution bands. It not really ‘restores’ any information, because it was simply not measured by the sensor. My advice is to use the S2 Resampling operator to bring all bands to 10m and apply C2RCC on the result. This works based on my experience.

#3

HI ABraun,

I tried super resolution because I thought after applying C2RCC on the upsampled images, there will be less variation in the Rrs across pixels. The main aim is to estimate Chl-a.

I had also tried S2 resampling but there is high variation in Rrs across the pixels. If I upsample from 60m to 10m, the variation in Rrs and Chl-a across the pixels should be less than 60 m images. But S2 resampling has increased the variation.

#4

sorry, what does “Rrs” stand for?
Maybe you can give examples why the S2 resampling produces results which do not match your requirements (the variation you report).

#5

Remote sensing reflectance.

#6

If you aim at less variation in the reflectance, you could simply resample to 10 meters using the nearest neighbor method, which simply maintains the values from the 20 and 60 meter bands in the contained 10 meter pixels.

#7

Yes I have selected the nearest beighbor method.

<node id="S2Resampling">
    <operator>S2Resampling</operator>
    <sources>
      <sourceProduct refid="Read"/>
    </sources>
    <parameters>
      <resolution>10</resolution>
      <upsampling>Nearest</upsampling>
      <downsampling>Mean</downsampling>
      <flagDownsampling>First</flagDownsampling>
      <resampleOnPyramidLevels>true</resampleOnPyramidLevels>
    </parameters>
#8

Then I don’t understand your problem :slight_smile:

#9

I’ you mean by ‘varaition’ that the images look noisy then this is a known issue for C2RCC and S2 data.
This is under investigation. However, reasmpling shouldn’t change this much.