After noise reduction some images are fully black (zero).
This is caused by zero values in the source product. These zero values cause NaN values when destriping factors are computed. And those NaN values cause the wrong zeros in the radiance bands.
The zero values in the chris source products were only observed in recent products. For example, in CHRIS_U7_201016_731E_41.hdf radiance_1 has zero values
The first question is if these values should be considered as no-data in the source product. The format specification says:
“PPU reset errors have been set to zero (originally resulted in negative values in previous releases of the data.)”
The second question is how these values should be treated in the algorithm.
The original algorithm implementor suggested to replace the zero values by one during the computation of the destriping factors. The exact value is not important here and one is close enough to zero compared to valid values. The result using this approach looks reasonable.