Feature request: mask dilation or pixel based distance

[alex-s-gardner]

The Images package has a really powerful pixel distance calculation (similar to bwdist in matlab: https://www.mathworks.com/help/images/ref/bwdist.html)... I use it all the time in my work and suspect other would appreciate it as well..

Right now I do this:

Images.distance_transform(feature_transform(BitMatrix)) 

which return the pixel distance matrix.

I'm wondering if we should just stick with this or if there is a good reason to provide better integration between Rasters and Images such that users could do things like:

raster_mask_dialated = Rasters.dialate(raster_mask, 2)
raster_distance_from_mask = Rasters.pixeldist(raster_mask)

Not sure what the philosophy here is.

[rafaqz]

Why not add those functions to Images.jl for any array?

What's the advantage of Rasters defining them?

[alex-s-gardner]

This functions (capabilities) are in Images and can be preformed on any array.

I guess I was simply thinking that helper functions within Rasters would improve user experience as it's a bit of a lift to learn a new package to do what I would consider a common operation of mask dilation. But I also understand that it might not makes sense, as it adds complexity to Rasters and another dependency.. maybe it also goes against the ethos of compostability.

[rafaqz]

Yeah if it already works fine, it just needs documentation so people know they can do it. And maybe better syntax in Images.jl.

The case for adding it here is where it doesn't return a Raster, but should

[alex-s-gardner]

Going to leave this open to motivate the generation of an example.. and maybe improved Images.jl syntax

[rafaqz]

Yes maybe we can get a longer list of Images.jl things that should work here and put them in docs and testing, then push any syntax changes we want back to Images.jl