Paganin filter: padding_method options
#245
Conversation
| distance: float = 1.0, | ||
| energy: float = 53.0, | ||
| ratio_delta_beta: float = 250, | ||
| padding_method: Union[ |
There was a problem hiding this comment.
I think padding_method parameter can be slightly confusing to users. I would think that the padding method is something how you deal with the boundary, e.g. edge/symmetric/reflect etc. I believe we do not expose such option for now and by default it is always edge . Here it is how the padding value is calculated. So may be this parameter should be called calculate_padding_value_method instead? But then, we also allow to provide an integer to it, which makes it slightly ambiguous.
Also the padding integer should be a list, separated for X and Y dimensions. So may be we should remove this option for now and provide only "next_power_of_2", "next_fast_length" ? Hope it makes sense?
|
|
||
| cp = cupywrapper.cp | ||
| cupy_run = cupywrapper.cupy_run | ||
| next_fast_len = cupywrapper.next_fast_len |
There was a problem hiding this comment.
I'd rather make an import from scipy here, cupywrapper is mostly to deal with the CuPy related imports
Added
padding_methodparameter topaganin_filter. Options:next_power_of_2(default, original method): Pads data to the size that is the next power of 2next_fast_length: Pads data to the next size that is fast for FFT. This can be zero padding. Uses less memory thannext_power_of_2Updated tests and zenodo tests.
Fixes: Paganin filter padding options