Given a 3D object with principal moments V_0 <= V_1 <= V_2, itkShapeLabelMapFilter computes elongation as sqrt(V_2 / V_1).
It doesn't compute eccentricity, so I had added that to LabelGeometryMeasures as sqrt(1 - V_2^2 / V_0^2).
But I think it might be better to use sqrt(1 - V_2^2 / V_1^2), so that it works on the same moments as the elongation. For cortex and other thin structures, the relative shape of the longer sides are often more interesting anyway.
The new filter also has roundness, which is a measure of similarity to a sphere and considers all three axes.
Related to #1733
Given a 3D object with principal moments V_0 <= V_1 <= V_2, itkShapeLabelMapFilter computes elongation as sqrt(V_2 / V_1).
It doesn't compute eccentricity, so I had added that to LabelGeometryMeasures as sqrt(1 - V_2^2 / V_0^2).
But I think it might be better to use sqrt(1 - V_2^2 / V_1^2), so that it works on the same moments as the elongation. For cortex and other thin structures, the relative shape of the longer sides are often more interesting anyway.
The new filter also has roundness, which is a measure of similarity to a sphere and considers all three axes.