Closed koegl closed 5 months ago
No, there's currently no function to transform landmarks directly.
order=0
, and look at the results. That would at least give you a visual confirmation.You could also have a look at the Python implementation that is used for the tests: https://github.com/gvtulder/elasticdeform/blob/master/test_deform_grid.py#L26-L62
That function computes the coordinates map for the output image. For example, coordinates[10, 13] = [9, 12]
would mean that pixel (10, 13) in the output is based on pixel (9, 12) of the input.
In a sense, it's the opposite of what you want: if you know where the landmark is in the output, you can look up where it was in the input. You might be able to use this in reverse, but I don't know if that also works for your application.
If you figure out a way to compute it, it would be a nice thing to add.
I'm working on image registration, where I have a moving image that I want to register to a fixed image.
Both images have a set of corresponding landmarks, that I need to use to evaluate the quality of the registration.
To do this, I need to deform the moving image and the corresponding landmarks with the same deformation field.
Is there a way to use Elasticdeform to deform a set of landmarks (i.e. points with some coordinates)?