/* For the _fast_ implementation this could happen:
* Infinity could happen if only the numerator is 0, i.e.:
* sum1squared == sum1 * sum1 * invN
* and
* sum2squared == sum2 * sum2 * invN
* If the denominator is also zero, one will get NaN, i.e:
* sumProduct1_2 == sum1 * sum2 * invN
*
* For the classic implementation it could happen, too:
* Infinity happens if one channels sum of value-mean-differences
* is zero. If it is negative for one image you will get NaN.
* Additionally, if is zero for both channels at once you
* could get NaN. NaN
*/
thrownewMissingPreconditionException("A numerical problem occured: the input data is unsuitable for this algorithm. Possibly too few pixels (in range were: "+iterations+").");