Gregory Lee

Thanks @judafe, good to see you back here. I wouldn't worry about those warnings. I had been meaning to fix the I/O ones for awhile, so this was a good...

The form used in scikit-image is the simplified one shown in Eqn. 5 in [this open access IPOL article](https://www.ipol.im/pub/art/2018/196/). Ideally the two forms should be equivalent, but in practice there...

Another one is `measure.inertia_tensor_eigvals` which should be `measure.inertia_tensor_eigenvalues`

If we are going to rename to `hessian_tensor_eigenvalues` as opposed to `hessian_matrix_eigenvalues` then we should probably also rename `hessian_matrix`->`hessian_tensor` for consistency.

Thanks @CSSFrancis! I will try to review this in the coming week.

Here is the output of the demo showing that the side of the eagle that was mostly in shadow becomes much more visible after filtering: 

Sorry it took so long to get back to this one. Thanks for reviewing @mkcor and @alexdesiqueira, your suggestions have been applied.

> This is strange. Although using masked arrays works in some cases, when trying to corner the problem I found this minimal Thanks, @plu2k, I can confirm that this is...

There are n-dimensional distance transforms for binary images in `scipy.ndimage.distance_transform_edt` and `scipy.ndimage.distance_transform_cdt`. Do either of those work for your use case?

## current main branch ``` benchmarks_moments.MomentsSuite.time_moments_central -- order --------------------------------- -------------------------------------- shape dtype 1 2 3 ================= =============== ============ ============ ============ (64, 64) numpy.uint8 62.8±0.5μs 62.7±0.4μs 62.9±0.2μs (64, 64) numpy.float32 51.0±0.4μs...