Move Shepp-Logan phantom to TestImages.jl?

[stillyslalom]

It seems a bit out of place in algorithms.jl.

Images.jl/src/algorithms.jl

Line 598 in 7bdd97d

function shepp_logan(M,N; highContrast=true)

[johnnychen94]

A good idea to me. I've thought about this once a while, but just can't come up with a compatible API to the current testimage usage.

Probably would just move them around.

CRef: JuliaImages/ImageReconstruction.jl#4

[stillyslalom]

If there's not a good way to shoehorn it into a testimage function call, TestImages.jl can easily have more than one exported method.

[johnnychen94]

I'm good to this, mind to make a PR?

Examples can be found at JuliaImages/ImageMorphology.jl#28 and #895

Doing this would introduce TestImages as a dependency to Images.jl, does this sound good to you? @timholy (Or just add a deprecation warning about deleting it in Images.jl?)

Instead of reexport TestImages.shepp_logan in Images.jl, we should just add a deleting warning about this, and redirect users to TestImages.

[timholy]

I like the idea of moving it. We would have to do this in a way that is non-breaking, though, unless we bump minor or major versions.

[johnnychen94]

For clarification, I was suggesting (1) copying it to TestImages, (2) add a deprecation in Images.jl, and finally (3) deleting it in Images@v1.0.0

[johnnychen94]

One way to make it compatible with the current testimage API is to redirect testimage("phantom", size=(256, 256), high_contrast = true) to its real generating function.

Note that "phantom" is the fullname of this image. No extension required.

[johnnychen94]

Some update on this:

shepp_logan is reimplemented in TestImages.shepp_logan (JuliaImages/TestImages.jl#77) and Images.shepp_logan will be removed in a future release (possibly v1.0.0)

[timholy]

This was done for the 0.25 release.