Hi @pitkajuh,

Thanks for tackling this bug! Please update the tests accordingly and document the changes in comments, e.g., by referring to the SciPy implementation (as done in #7357).

Hi @pitkajuh,

Thanks for tackling this bug! Please update the tests accordingly and document the changes in comments, e.g., by referring to the SciPy implementation (as done in #7357).

Hello. Yes, sorry about that. I fixed the tests and now they should work. The problem was just as described in #7357. Removing and adding minus signs from/to appropriate places fixed the issue and now the result are identical to the SciPy implementation.

@lagru since we are changing the return value of a function, I think we should deprecate this function and replace it with, say, ski.filters.laplacian so that users know they should update/check current code with ski.filters.laplace(arr). What do you think?

With the discussion in #7357 (comment), I'm not convinced that we changing the definition is worth it.

I don't like the suggestion to sidestep the issue by renaming to laplacian. It seems to me that the term "laplace filter" is more common, so I'd prefer to keep skimage.filters.laplace around and instead document the definition and how it differs to SciPy in the docstring of both functions.

That said, I could live with us switching the definition for skimage2.