I have a 3D mask volume L with the following:
print(L.shape)
(170, 256, 256)
print("L: ", np.unique(L))
L: [0 1 2 3 4 5 6 7 8]
I want to downsample and then upsample the mask to its original size keeping the label values the same.
Failed attempt->Downsampling:
from scipy.ndimage.interpolation import zoom
zL = zoom(L, (0.5, 0.5, 0.5), mode='nearest')
print("zL: ", np.unique(zL))
zL: [-3 -2 -1 0 1 2 3 4 5 6 7 8 9 10]
which is changing the label values.
I have tried with other mode options such as constant and all. But none seems to work.
Successful attempt->Downsampling: On the other hand:
dx = 2
dy = 2
dz = 2
if DOWNSAMPLE:
L_down = L[::dx, ::dy, ::dz]
print("L_down: ", np.unique(L_down))
L_down: [0 1 2 3 4 5 6 7 8]
seems to work without changing the label values.
Failed attempt->Upsampling:
But when going back to original size with
zL = zoom(L_down, (2, 2, 2), mode='nearest')
did not work.
Also if there are any other details about trilinear interpolation or so would be appreciated.
Thanks.
