Seems like this resource allows someone to apply a transformation matrix (I don’t see a way to solve for one). So I can write the application function like:
using ImageTransformations
transform_perspective(m::Matrix, point::Vector) = PerspectiveMap() ∘ inv(LinearMap(m)) ∘ \x -> push(x, 1)
I’m not clear what I’m gaining compared to the simpler code above. Isn’t this a lot more operations to create all the transformation objects?