Dataset of crops extracted from real L1A data.
This dataset is composed of 940 stacks of 15 256x256 images each.
A stack is produced by registering the 15 frames with integer translations. This means that the frames are purposefully not exactly aligned since it would require resampling of the data.
We calibrated the radiometry in order to stabilize the intensities.
A 256x256 pixel crop is then extracted.