Closed jni closed 13 years ago
Surface between two n-dimensional objects is (n-1)-dimensional. (?) Now, to find the largest plane to have the same bounding box as the line, we sort the dimensions in increasing order (a, b, c), then take sqrt(a^2+b^2) * c. However, in digital space the plane will be approximated and so it suffices to take b*c, or, in general:
a.shape.sort()[1:].prod() * boundary_thickness
boundary_thickness
is 1 for standard watershed, 2 for damless watershed, or other for thicker boundaries.
Added as of 36894c36067963fe77a9701db3054cddb6789c8f.
Other managers may be possible in the future, however. Nathan Clack suggested moments (mean, variance, etc) of local curvature estimates. Curvature can be obtained as derivatives of normal vector along axes.
Proposed features: ratio of surface area vs bounding box surface area ratio of surface area vs maximum and/or minimum flat surface area within bounding box variance of curvature along surface