def test_get_img_transformation_matrix(self):
# simplest matrix
md = {
model.MD_PIXEL_SIZE: (1.0, 1.0),
model.MD_ROTATION: 0.0,
model.MD_SHEAR: 0.0,
}
mat = get_img_transformation_matrix(md)
numpy.testing.assert_almost_equal(mat, [[1., 0.], [0., -1.]])
micro = 1.0e-06
# 90 degrees of rotation
md = {
model.MD_PIXEL_SIZE: (micro, micro),
model.MD_ROTATION: math.pi / 2,
}
mat = get_img_transformation_matrix(md)
numpy.testing.assert_almost_equal(mat, [[0.0, micro], [micro, 0.0]])
# 180 degrees of rotation
md = {
model.MD_PIXEL_SIZE: (micro, micro),
model.MD_ROTATION: math.pi,
}
mat = get_img_transformation_matrix(md)
numpy.testing.assert_almost_equal(mat, [[-micro, 0.0], [0.0, micro]])
# scale and rotation 45 degrees
md = {
model.MD_PIXEL_SIZE: (micro, micro),
model.MD_ROTATION: math.pi / 4,
}
mat = get_img_transformation_matrix(md)
sin = math.sin(math.pi / 4) * micro
numpy.testing.assert_almost_equal(mat, [[sin, sin], [sin, -sin]])
# scale and rotation 45 degrees
md = {
model.MD_PIXEL_SIZE: (micro, micro),
model.MD_SHEAR: 0.5,
}
mat = get_img_transformation_matrix(md)
numpy.testing.assert_almost_equal(
mat, [[micro, 0.0], [-0.5 * micro, -micro]])
# everything
md = {
model.MD_PIXEL_SIZE: (micro, micro),
model.MD_ROTATION: math.pi / 4,
model.MD_SHEAR: 0.1,
}
mat = get_img_transformation_matrix(md)
numpy.testing.assert_almost_equal(
mat, [[7.77817459e-07, sin], [6.36396103e-07, -sin]])
# nothing
md = {}
with self.assertRaises(ValueError): # MD_PIXEL_SIZE must be present
mat = get_img_transformation_matrix(md)
def test_get_img_transformation_matrix(self):
# simplest matrix
md = {
model.MD_PIXEL_SIZE: (1.0, 1.0),
model.MD_ROTATION: 0.0,
model.MD_SHEAR: 0.0,
}
mat = get_img_transformation_matrix(md)
numpy.testing.assert_almost_equal(mat, [[1., 0.], [0., -1.]])
micro = 1.0e-06
# 90 degrees of rotation
md = {
model.MD_PIXEL_SIZE: (micro, micro),
model.MD_ROTATION: math.pi / 2,
}
mat = get_img_transformation_matrix(md)
numpy.testing.assert_almost_equal(mat, [[0.0, micro], [micro, 0.0]])
# 180 degrees of rotation
md = {
model.MD_PIXEL_SIZE: (micro, micro),
model.MD_ROTATION: math.pi,
}
mat = get_img_transformation_matrix(md)
numpy.testing.assert_almost_equal(mat, [[-micro, 0.0], [0.0, micro]])
# scale and rotation 45 degrees
md = {
model.MD_PIXEL_SIZE: (micro, micro),
model.MD_ROTATION: math.pi / 4,
}
mat = get_img_transformation_matrix(md)
sin = math.sin(math.pi / 4) * micro
numpy.testing.assert_almost_equal(mat, [[sin, sin], [sin, -sin]])
# scale and rotation 45 degrees
md = {
model.MD_PIXEL_SIZE: (micro, micro),
model.MD_SHEAR: 0.5,
}
mat = get_img_transformation_matrix(md)
numpy.testing.assert_almost_equal(mat, [[micro, 0.0], [-0.5 * micro, -micro]])
# everything
md = {
model.MD_PIXEL_SIZE: (micro, micro),
model.MD_ROTATION: math.pi / 4,
model.MD_SHEAR: 0.1,
}
mat = get_img_transformation_matrix(md)
numpy.testing.assert_almost_equal(mat, [[7.77817459e-07, sin], [6.36396103e-07, -sin]])
# nothing
md = {}
with self.assertRaises(ValueError): # MD_PIXEL_SIZE must be present
mat = get_img_transformation_matrix(md)
def getCenterOfTiles(tiles, result_shape):
""" Calculates the center of the result image
It is based on the formula for calculating the position of a pixel in world coordinates:
CT = CI + TMAT * DC
where:
CT: center of the tile in pixel coordinates
CI: center of the image in world coordinates
DC: delta of the centers in pixel coordinates
TMAT: transformation matrix
From the formula above, comes the following formula:
CI = CT - TMAT * DC,
which is used below
tiles (tuple of tuple of DataArray): Tiles
result_shape (height, width): Size in pixels of the result image from the tiles
return (x, y): Physical coordinates of the center of the image
"""
first_tile = tiles[0][0]
ft_md = first_tile.metadata
dims = ft_md.get(model.MD_DIMS, "CTZYX"[-first_tile.ndim::])
ft_shape = [
first_tile.shape[dims.index('X')], first_tile.shape[dims.index('Y')]
]
# center of the tile in pixel coordinates
center_tile_pixel = [d / 2 for d in ft_shape]
# center of the image in pixel coordinates
center_image_pixel = [d / 2 for d in result_shape[::-1]]
# distance between the center of the tile and the center of the image, in pixel coordinates
dist_centers_tile_pixels = [
ct - ci for ct, ci in zip(center_tile_pixel, center_image_pixel)
]
# converts the centers distance, so this variable can be multiplied by the transformation matrix
dist_centers_tile_pixels = numpy.array(
dist_centers_tile_pixels).transpose()
# transformation matrix
tmat = get_img_transformation_matrix(first_tile.metadata)
# distance of the centers converted to world coordinates
dist_centers_w = tmat @ dist_centers_tile_pixels
# convert the variable from a numpy.matrix to a numpy.array
dist_centers_w = numpy.ravel(dist_centers_w)
# center of the tile in world coordinates
center_tile_w = first_tile.metadata[model.MD_POS]
# center of the image in world coordinates
image_pos = center_tile_w - dist_centers_w
return tuple(image_pos)
def getCenterOfTiles(tiles, result_shape):
""" Calculates the center of the result image
It is based on the formula for calculating the position of a pixel in world coordinates:
CT = CI + TMAT * DC
where:
CT: center of the tile in pixel coordinates
CI: center of the image in world coordinates
DC: delta of the centers in pixel coordinates
TMAT: transformation matrix
From the formula above, comes the following formula:
CI = CT - TMAT * DC,
which is used below
tiles (tuple of tuple of DataArray): Tiles
result_shape (height, width): Size in pixels of the result image from the tiles
return (x, y): Physical coordinates of the center of the image
"""
first_tile = tiles[0][0]
ft_md = first_tile.metadata
dims = ft_md.get(model.MD_DIMS, "CTZYX"[-first_tile.ndim::])
ft_shape = [first_tile.shape[dims.index('X')], first_tile.shape[dims.index('Y')]]
# center of the tile in pixel coordinates
center_tile_pixel = [d / 2 for d in ft_shape]
# center of the image in pixel coordinates
center_image_pixel = [d / 2 for d in result_shape[::-1]]
# distance between the center of the tile and the center of the image, in pixel coordinates
dist_centers_tile_pixels = [ct - ci for ct, ci in zip(center_tile_pixel, center_image_pixel)]
# converts the centers distance, so this variable can be multiplied by the transformation matrix
dist_centers_tile_pixels = numpy.matrix(dist_centers_tile_pixels).getT()
# transformation matrix
tmat = get_img_transformation_matrix(first_tile.metadata)
# distance of the centers converted to world coordinates
dist_centers_w = tmat * dist_centers_tile_pixels
# convert the variable from a numpy.matrix to a numpy.array
dist_centers_w = numpy.ravel(dist_centers_w)
# center of the tile in world coordinates
center_tile_w = first_tile.metadata[model.MD_POS]
# center of the image in world coordinates
image_pos = center_tile_w - dist_centers_w
return tuple(image_pos)