It seems that it should be something simple-ish to implement. If you know that your expansion only is going to have terms so that m+l = even (or odd), it should be the case that just giving half the sample points, only in the upper sphere, should be enough to compute the transform. Especially since the recursions already separate both cases.
It seems that it should be something simple-ish to implement. If you know that your expansion only is going to have terms so that m+l = even (or odd), it should be the case that just giving half the sample points, only in the upper sphere, should be enough to compute the transform. Especially since the recursions already separate both cases.