Unlike the NaN values in floating point format, integers don't have any predefined defined blank value. So as explained in the blank pixels section of the Gnuastro book, we have adopted the following convention for integers: for signed integers, the blank value is the smallest (most negative) and for unsigned integers, its the maximum value.

For unsigned integers (which is usually for type for raw exposures), this is natural because the maximum value is a saturated pixel.

But in some scenarios, this is not the desired behavior and the minimum (for signed) or maximum (for unsigned) values actually have a meaning. We should define a framework to disable associating blank values to integers if the user asks. Maybe as a common option to all programs.