def rotaxis_rough_filt(proj, N_steps=10, sigma=5, accuracy=10):
"""calculate the rotation axis comparing 0 and 180 projection shift
Parameters
__________
proj: 3D array
N_steps: projections per degree
by default it compares the central part of images (between 1/4 and 3/4 of shape)
"""
a = proj.shape[1] // 4
b = 3 * proj.shape[1] // 4
c = proj.shape[2] // 4
d = 3 * proj.shape[2] // 4
cent = []
N_rot = proj.shape[0] - 180 * N_steps
for i in range(N_rot):
im1 = filt_gauss_laplace(proj[i, a:b, c:d], sigma)
im2 = np.flip(
filt_gauss_laplace(proj[i + N_steps * 180, a:b, c:d], sigma), 1)
distances = shift(im1, im2, accuracy)
cent.append(proj[i].shape[1] / 2 + distances[1] / 2)
return cent
def read_flat(j):
"""
j: int
an index of the file that should be processed
Please note, j always starts from zero
To open correct file, images array uses images[i][j + N_start-1]
"""
#global ff_shared, ff_con, shifts, indexes, Pro
#set local variables
ff = tonumpyarray(ff_shared.shared_array_base, ff_shared.shape,
ff_shared.dtype)
proj_loc = tonumpyarray(proj.shared_array_base, proj.shape, proj.dtype)
shift = tonumpyarray(shifts.shared_array_base, shifts.shape, shifts.dtype)
#LoI = tonumpyarray(indexes.shared_array_base, indexes.shape, indexes.dtype)
ROI_ff = Pro.ROI_ff
ROI = Pro.ROI
images = Pro.images
N_start = Pro.N_start
Npad = Pro.Npad
#read image and do ff-retrieval
filt = []
for i in np.arange(len(images)):
im = tifffile.imread(images[i][j + N_start - 1])[ROI[1]:ROI[3],
ROI[0]:ROI[2]]
index = SSIM_sf.SSIM(
SSIM_const(im[ROI_ff[1]:ROI_ff[3], ROI_ff[0]:ROI_ff[2]]),
ff_con[i]).ssim()
im = im / ff[i][np.argmax(index)]
filt.append(im)
#LoI[j,i] = np.argmax(index)
#calculate shift for holograms
im_gau0 = var.filt_gauss_laplace(filt[0][ROI_ff[1]:ROI_ff[3],
ROI_ff[0]:ROI_ff[2]],
sigma=5)
for i in np.arange(len(filt)):
im_gau1 = var.filt_gauss_laplace(filt[i][ROI_ff[1]:ROI_ff[3],
ROI_ff[0]:ROI_ff[2]],
sigma=5)
shift[j, i] = (var.shift_distance(im_gau0, im_gau1, 10))
#shift images
filt = multiCTF.shift_imageset(np.asarray(filt), shift[j])
filt = np.asarray(filt)
#do CTF retrieval
filt = np.pad(filt, ((0, 0), (Npad, Npad), (Npad, Npad)),
'edge') # padding with border values
filt = multiCTF.multi_distance_CTF(filt, beta_delta, fresnelN,
zero_compensation)
filt = filt[Npad:(filt.shape[0] - Npad), :] # unpad images from the top
#rotate the image to compensate for the inclined rotation axis
#filt = rotate(filt, inclination, mode = 'nearest', axes = (1,0))
#save to memory
proj_loc[j] = filt