The problem with your example is probably the d parameter.
With d=1-a-b-c no scaling happens. But your d value is different, so you request an additional scaling. With the additional scaling the corrected image remains inside the initial image rectangle and there is no need for an additional rescaling. So the output of fulla with and without is the same.
The switch -s works, but not as the thread opened is wishing.
fulla -s -g -0.004318: 0.039896: -0.096925: 1.061347 -o a.tif input.tif 0.039896: -0.096925: 1.061347 -o b.tif input.tif a213b8222f80550 26 a.tif 875e20011457ecf b3 b.tif
Saving a.tif
fulla -g -0.004318:
Saving b.tif
md5sum a.tif b.tif
492bc5efef088c6
31e68917a31fd77
The problem with your example is probably the d parameter.
With d=1-a-b-c no scaling happens. But your d value is different, so you request an additional scaling. With the additional scaling the corrected image remains inside the initial image rectangle and there is no need for an additional rescaling. So the output of fulla with and without is the same.