def batch_displacement_warp3d(imgs, vector_fields):
"""
warp images by displacement vector fields
Parameters
----------
imgs : tf.Tensor
images to be warped
[n_batch, xlen, ylen, zlen, n_channel]
vector_fields : tf.Tensor
[n_batch, 3, xlen, ylen, zlen]
Returns
-------
output : tf.Tensor
warped imagees
[n_batch, xlen, ylen, zlen, n_channel]
"""
n_batch = tf.shape(imgs)[0]
xlen = tf.shape(imgs)[1]
ylen = tf.shape(imgs)[2]
zlen = tf.shape(imgs)[3]
grids = batch_mgrid(n_batch, xlen, ylen, zlen)
T_g = grids + vector_fields
output = batch_warp3d(imgs, T_g)
return output
def batch_affine_warp3d(imgs, theta):
"""
affine transforms 3d images
Parameters
----------
imgs : tf.Tensor
images to be warped
[n_batch, xlen, ylen, zlen, n_channel]
theta : tf.Tensor
parameters of affine transformation
[n_batch, 12]
Returns
-------
output : tf.Tensor
warped images
[n_batch, xlen, ylen, zlen, n_channel]
"""
n_batch = tf.shape(imgs)[0]
xlen = tf.shape(imgs)[1]
ylen = tf.shape(imgs)[2]
zlen = tf.shape(imgs)[3]
theta = tf.reshape(theta, [-1, 3, 4])
matrix = tf.slice(theta, [0, 0, 0], [-1, -1, 3])
t = tf.slice(theta, [0, 0, 3], [-1, -1, -1])
grids = batch_mgrid(n_batch, xlen, ylen, zlen)
grids = tf.reshape(grids, [n_batch, 3, -1])
T_g = tf.batch_matmul(matrix, grids) + t
T_g = tf.reshape(T_g, [n_batch, 3, xlen, ylen, zlen])
output = batch_warp3d(imgs, T_g)
return output