Get exact location of sampling grid points

[gfotedar]

Is there a way to get the exact location of sampling grid points so I can display the grid before and after BSpline deformation superimposed on the image?

[keppenhof]

Yes, the Grid class has a grid attribute that has the coordinates scaled between zero and one. You can multiply them by the size of the image using the scaled_to() method to get the exact locations:

grid = gryds.Grid(image_shape)
transformed_grid = grid.transform(some_transform, ...)
resized_grid = transformed_grid.scaled_to(image_shape)

The locations in image coordinates are in resized_grid.grid

[gfotedar]

That is not actually what I am looking for. Like in your example on your home page, in this part of code:

# Define a random 3x3 B-spline grid for a 2D image:
random_grid = np.random.rand(2, 3, 3)
random_grid -= 0.5
random_grid /= 5

You define a random_grid of size 3x3. I am looking for locations of these 9 points in [row,column] format, before and after the deformation. I can find their locations after the deformation, using what you told me above, but I need to know the locations before deformation to do that.

[keppenhof]

If I understand you correctly you are looking for the coordinates of the B-spline transformation's control points before the transformation.

These are never explicitly computed in the code. Instead the grid (like the random_grid in the code you reference) is resized to the size of the image you want to transform using B-spline interpolation.

Therefore, you can assume the B-spline control points are evenly distributed over the image's domain, i.e. in 2D

• If you have a 1x1 grid of control points, the single control point will be in the middle of the image;
• If you have a 2x2 grid of control points, each of the control points will be centered in a quarter of the image, i.e. if you have a 6x6 image, the control points will be at (1, 1), (1, 4), (4, 1), and (4, 4);
• If you have a 3x3 grid of control points, each of the control points will be centered in a ninth of the image, i.e. if you have a 9x9 image, the control points will be at (1, 1), (1, 4), (1, 7), (4, 1), (4, 4), (4, 7) etc.