def find_center_blurred(img, msk, x0=0, y0=0, threshold=None, blur_radius=4., dmax=5):
"""
Find the center using blurred version of 'pixelwise' method.
usage:
======
x,y = find_center_blurred(img, msk, x0=0, y0=0, threshold=None, blur_radius=4, dmax=5)
"""
if threshold is None: threshold = img.min()
I = spimage.sp_image_alloc(img.shape[1],img.shape[0],1)
I.image[:] = img * (img >= threshold)
I.mask[:] = msk[:]
I.detector.image_center[:] = np.array([center_to_pos(x0,img.shape[1]),
center_to_pos(y0,img.shape[0]), 0 ])
kernel = spimage.sp_gaussian_kernel(float(blur_radius),int(blur_radius*8+1),int(blur_radius*8+1),1)
c = spimage.sp_image_convolute_with_mask(I,kernel,np.array([1,1,1]).astype(np.int32))
ds = spimage.sp_image_alloc(img.shape[1]/4,img.shape[0]/4,1)
ds.image[:] = c.image[:-3:4,:-3:4]
ds.mask[:] = c.mask[:-3:4,:-3:4]
ds.detector.image_center[:] = c.detector.image_center[:] / 4.0
spimage.sp_find_center_refine_minimal_mask(ds, int(1+dmax/4), 0)
c.detector.image_center[:] = ds.detector.image_center[:] * 4.0
spimage.sp_image_free(ds)
score = spimage.sp_find_center_refine_minimal_mask(c, 4, 0)
x = pos_to_center(c.detector.image_center[0],img.shape[1])
y = pos_to_center(c.detector.image_center[1],img.shape[0])
spimage.sp_image_free(I)
spimage.sp_image_free(kernel)
spimage.sp_image_free(c)
return (x,y,score)
def allocate_image(shape):
"""Allocate spimage image object from numpy type shape"""
if len(shape) == 2:
img = _spimage.sp_image_alloc(shape[0], shape[1], 1)
elif len(shape) == 3:
img = _spimage.sp_image_alloc(shape[0], shape[1], shape[2])
else:
raise ValueError("Array must be 2 or 3 dimensional")
return img
def allocate_image(shape):
"""Allocate spimage image object from numpy type shape"""
if len(shape) == 2:
img = _spimage.sp_image_alloc(shape[0], shape[1], 1)
elif len(shape) == 3:
img = _spimage.sp_image_alloc(shape[0], shape[1], shape[2])
else:
raise ValueError("Array must be 2 or 3 dimensional")
return img
def numpy_array_to_image(img, msk=None):
s = img.shape
d = len(list(s))
if d == 3:
sp_img = spimage.sp_image_alloc(s[2], s[1], s[0])
else:
sp_img = spimage.sp_image_alloc(s[1], s[0], 1)
sp_img.image[:] = img[:]
if msk is not None:
sp_img.mask[:] = msk[:]
return sp_img
def numpy_array_to_image(img,msk=None):
s = img.shape
d = len(list(s))
if d == 3:
sp_img = spimage.sp_image_alloc(s[2],s[1],s[0])
else:
sp_img = spimage.sp_image_alloc(s[1],s[0],1)
sp_img.image[:] = img[:]
if msk is not None:
sp_img.mask[:] = msk[:]
return sp_img
def _init_initial_support(self):
if not self._initial_support_dirty:
self._log("Initial support already initialised.","DEBUG")
return
for img in [self._sp_initial_support,self._sp_initial_support_sh]:
if img is not None:
spimage.sp_image_free(img)
self._sp_initial_support = spimage.sp_image_alloc(self._Ny,self._Nx,1)
if "radius" in self._initial_support_config or "area" in self._initial_support_config:
X,Y = np.meshgrid(np.arange(self._Nx),np.arange(self._Ny))
X = X-(self._Nx-1)/2.
Y = Y-(self._Ny-1)/2.
R = np.sqrt(X**2 + Y**2)
self._phaser_dirty = True
if "radius" in self._initial_support_config:
r = self._initial_support_config["radius"]
else:
r = np.sqrt( self._initial_support_config["area"] * self._Nx * self._Ny / np.pi )
self._sp_initial_support.image[:] = np.float32(np.fft.fftshift(R) < r)
else:
self._phaser_dirty = True
S = self._initial_support_config["support_mask"]
self.sep_initial_support.image[:] = np.float32(S)
self._initial_support_dirty = False
self._log("Initial support initialised.","DEBUG")
def split_pnccd(filename):
f = h5py.File(filename)
i1 = f.keys().index('data')
i2 = f.values()[i1].keys().index('data1')
i2 = f.values()[i1].keys().index('data1')
data2 = f.values()[i1].values()[i2].value
data2_1 = spimage.sp_image_alloc(data2.shape[1]/2,data2.shape[0],1)
data2_2 = spimage.sp_image_alloc(data2.shape[1]/2,data2.shape[0],1)
data2_1.image[:,:] = data2[:,:(data2.shape[1]/2)]
data2_2.image[:,:] = data2[:,(data2.shape[1]/2):]
spimage.sp_image_write(data2_1,filename[:-3]+"_part1.h5",0)
spimage.sp_image_write(data2_2,filename[:-3]+"_part2.h5",0)
def slice_3D(fn, output_dir=''):
#Takes 3 slices through the center along x,y and z from a 3D image and saves it as three new 2D images.
img_3D = spimage.sp_image_read(fn, 0)
dim = shape(img_3D.image)[0]
img_2D = spimage.sp_image_alloc(dim, dim, 1)
img_2D.shifted = img_3D.shifted
img_2D.scaled = img_3D.scaled
if img_3D.shifted == 0:
s = dim / 2
else:
s = 0
img_2D.image[:, :] = img_3D.image[s, :, :]
img_2D.mask[:, :] = img_3D.mask[s, :, :]
spimage.sp_image_write(img_2D,
output_dir + fn.split('.')[0] + '_x_slice.h5', 0)
img_2D.image[:, :] = img_3D.image[:, s, :]
img_2D.mask[:, :] = img_3D.mask[:, s, :]
spimage.sp_image_write(img_2D,
output_dir + fn.split('.')[0] + '_y_slice.h5', 0)
img_2D.image[:, :] = img_3D.image[:, :, s]
img_2D.mask[:, :] = img_3D.mask[:, :, s]
spimage.sp_image_write(img_2D,
output_dir + fn.split('.')[0] + '_z_slice.h5', 0)
def intensities_array_to_h5(img0,msk0,cx=None,cy=None,cropLength=None,save_to_file=None):
import spimage,imgtools
Nx = img0.shape[1]
Ny = img0.shape[0]
img = img0
msk = msk0
if cropLength != None:
img = imgtools.crop(img0,cropLength,center=[cy,cx])
msk = imgtools.crop(msk0,cropLength,center=[cy,cx])
Nx = cropLength
Ny = cropLength
elif cx != None and cy != None:
img = imgtools.recenter(img0,cx,cy)
msk = imgtools.recenter(msk0,cx,cy)
I = spimage.sp_image_alloc(Nx,Ny,1)
I.image[:,:] = img[:,:]
I.mask[:,:] = msk[:,:]
I2 = spimage.sp_image_shift(I)
I2.shifted = 0
I2.phased = 0
spimage.sp_image_free(I)
if save_to_file != None:
spimage.sp_image_write(I2,save_to_file,0)
else:
return I2
def crop_image(in_file, out_file, side, center=None):
"""Function to crop an h5 image and pad with zeros around it"""
img = spimage.sp_image_read(in_file, 0)
if not center:
center = img.detector.image_center[:2]
shifted = 0
if img.shifted:
shifted = 1
img = spimage.sp_image_shift(img)
cropped = spimage.sp_image_alloc(side, side, 1)
cropped.image[:, :] = image_manipulation.crop_and_pad(
img.image, center, side)
cropped.mask[:, :] = image_manipulation.crop_and_pad(
img.mask, center, side)
if shifted:
cropped = spimage.sp_image_shift(cropped)
spimage.sp_image_write(cropped, out_file, 16)
spimage.sp_image_free(img)
spimage.sp_image_free(cropped)
print("end")
def image_from_array(image, mask=None):
"""Create spimage image type from numpy array with optional mask"""
if len(image.shape) == 2:
img = _spimage.sp_image_alloc(image.shape[0], image.shape[1], 1)
elif len(image.shape) == 3:
img = _spimage.sp_image_alloc(image.shape[0], image.shape[1], image.shape[2])
else:
raise ValueError("Array must be 2 or 3 dimensional")
img.image[:] = image
if mask is not None:
img.mask[:] = _numpy.int32(mask)
else:
img.mask[:] = 1
img.shifted = 0
img.phased = int(bool(_numpy.iscomplex(image).sum() > 0))
return img
def image_from_array(image, mask=None):
"""Create spimage image type from numpy array with optional mask"""
if len(image.shape) == 2:
img = _spimage.sp_image_alloc(image.shape[0], image.shape[1], 1)
elif len(image.shape) == 3:
img = _spimage.sp_image_alloc(image.shape[0], image.shape[1],
image.shape[2])
else:
raise ValueError("Array must be 2 or 3 dimensional")
img.image[:] = image
if mask is not None:
img.mask[:] = _numpy.int32(mask)
else:
img.mask[:] = 1
img.shifted = 0
img.phased = int(bool(_numpy.iscomplex(image).sum() > 0))
return img
def _init_amplitudes(self):
if not self._amplitudes_dirty:
self._log("Amplitudes already initialised.","DEBUG")
return
A = self._amplitudes
if len(A.shape) == 2:
self._Nx = A.shape[1]
self._Ny = A.shape[0]
if self._mask is not None:
M = self._mask.copy()
else:
M = np.ones(shape=A.shape,dtype="bool")
for img in [self._sp_amplitudes]:
if img is not None:
spimage.sp_image_free(img)
self._sp_amplitudes = spimage.sp_image_alloc(A.shape[0],A.shape[1],1)
self._sp_amplitudes.image[:,:] = np.float32(A[:,:])
self._sp_amplitudes.mask[:,:] = np.int32(M[:,:])
self._sp_amplitudes.scaled = 1
self._sp_amplitudes.phased = 0
self._amplitudes_dirty = False
self._log("Amplitudes initialised.","DEBUG")
if len(A.shape) == 3:
self._Nx = A.shape[2]
self._Ny = A.shape[1]
self._Nz = A.shape[0]
if self._mask is not None:
M = self._mask.copy()
else:
M = np.ones(shape=A.shape,dtype="bool")
for img in [self._sp_amplitudes]:
if img is not None:
spimage.sp_image_free(img)
self._sp_amplitudes = spimage.sp_image_alloc(A.shape[0],A.shape[1],A.shape[2])
self._sp_amplitudes.image[:,:,:] = np.float32(A[:,:,:])
self._sp_amplitudes.mask[:,:,:] = np.int32(M[:,:,:])
self._sp_amplitudes.scaled = 1
self._sp_amplitudes.phased = 0
self._amplitudes_dirty = False
self._log("Amplitudes initialised.","DEBUG")
def pnccd_to_image(infile, outfile):
try:
f = h5py.File(infile)
except:
raise IOError("Can't read %s. It may not be a pnCCD file." % filename)
i1 = f.keys().index("data")
i2 = f.values()[i1].keys().index("data1")
data = f.values()[i1].values()[i2].value
img = spimage.sp_image_alloc(pylab.shape(data)[0], pylab.shape(data)[1], 1)
img.image[:, :] = data[:, :]
spimage.sp_image_write(img, outfile, 0)
spimage.sp_image_free(img)
def mask_center(img_file_name, radius, output_file_name=None, save_file=True):
"""Create a new mask around the center with given radius"""
img = spimage.sp_image_read(img_file_name, 0)
r_array = stuff.r_array_3D(img.image.shape[0])
img.mask[r_array < radius] = 0
img.mask[r_array > radius] = 1
new_mask = spimage.sp_image_alloc(*np.shape(img.mask))
new_mask.mask[:, :, :] = img.mask
new_mask.image[:, :, :] = img.image
new_mask.shifted = img.shifted
new_mask.scaled = img.scaled
new_mask.detector = img.detector
if save_file:
spimage.sp_image_write(new_mask, output_file_name, 0)
else:
return (new_mask)
def downsample_by_average(image, file_type, sf):
"""image can be either spimage object or numpy array, file_type = spimage_object or numpy_array, sf is the downsampling factor"""
if file_type == 'spimage_object':
model_side = shape(image.image)[0]
if file_type == 'numpy_array':
model_side = shape(image)[0]
downsampled = spimage.sp_image_alloc(model_side / sf, model_side / sf, 1)
for i, x in enumerate(range(0, model_side, sf)):
for j, y in enumerate(range(0, model_side, sf)):
if file_type == 'spimage_object':
downsampled.image[i, j] = average(image.image[x:x + sf,
y:y + sf])
downsampled.mask[i, j] = int((image.mask[x:x + sf,
y:y + sf]).all())
if file_type == 'numpy_array':
downsampled.image[i, j] = average(image[x:x + sf, y:y + sf])
return downsampled
def add_mask(path, number_of_files, output_file):
"""Reads in files in paths, outputs h5 file output_file with accumulative mask from input images saved in both image and mask"""
imgs = [
os.path.join(path, i) for i in os.listdir(path) if i.endswith('.h5')
]
for img_file in imgs[:number_of_files]:
img = spimage.sp_image_read(img_file, 0)
try:
msk_array = msk_array + img.mask
except NameError:
msk_array = img.mask
msk_array[msk_array < msk_array.max()] = 0
msk_array[msk_array != 0] = 1
new = spimage.sp_image_alloc(
numpy.shape(img.image)[0],
numpy.shape(img.image)[1], 1)
new.mask[:, :] = msk_array
new.image[:, :] = msk_array
spimage.sp_image_write(new, output_file, 0)
def find_center_pixelwise_fast(img, msk, x0=0, y0=0, dmax=5, rmax=None):
"""
Find center of diffraction pattern using a pixelwise comparison of centry-symmetric pixels.
This is a faster C implementation.
usage:
======
x,y = find_center_pixelwise_fast(img, msk, x0, y0, dmax=5, rmax=None)
"""
if rmax is not None: msk &= (spimage.rgrid(msk.shape, (x0,y0)) < rmax)
I = spimage.sp_image_alloc(int(np.ceil(img.shape[1])), int(np.ceil(img.shape[0])), 1)
I.image[:] = img
I.mask[:] = msk
I.detector.image_center[:] = np.array([center_to_pos(x0,img.shape[1]), center_to_pos(y0,img.shape[0]), 0 ])
success = spimage.sp_find_center_refine(I, dmax, 0, None)
x = pos_to_center(I.detector.image_center[0],img.shape[1])
y = pos_to_center(I.detector.image_center[1],img.shape[0])
spimage.sp_image_free(I)
return (x,y,success)
def calc_average_img(r2, r1=0, r_skip=None, iteration=None):
rundir_range = range(r1, r2)
if r_skip != None:
for i in r_skip:
rundir_range.remove(i)
fail = 0
for r in rundir_range:
run_dir = 'run_%04d' % r
print run_dir
if iteration == None:
iteration = return_last_iteration_integer(
'{r}/output_mnt/'.format(r=run_dir))
with h5py.File('{r}/output_mnt/fmodel_{i}.h5'.format(
r=run_dir, i=iteration)) as f:
try:
added_real += f['real'][...]
added_imag += f['imag'][...]
except NameError:
added_real = f['real'][...]
added_imag = f['imag'][...]
except KeyError:
print 'fail'
fail += 1
if r == r2 - 1:
mask = f['mask'][:]
added_real /= float(len(rundir_range) - fail)
added_imag /= float(len(rundir_range) - fail)
complex_fmodel = added_real + 1j * added_imag
new = spimage.sp_image_alloc(*np.shape(added_real))
new.phased = 1
new.mask[:] = mask
new.image[:] = complex_fmodel
spimage.sp_image_write(
new, 'avg_fmodel_runs_{i}_{j}_iteration{k}.h5'.format(i=r1,
j=r2,
k=iteration), 0)
new.image[:, :, :] = fft.fftn(complex_fmodel)
spimage.sp_image_write(
new, 'avg_model_runs_{i}_{j}_iteration{k}.h5'.format(i=r1,
j=r2,
k=iteration), 0)
def calc_average_img_translated(r2,
r1=0,
r_skip=None,
iteration=None,
reference_model=None):
rundir_range = range(r1, r2)
if r_skip != None:
for i in r_skip:
rundir_range.remove(i)
for r in rundir_range:
run_dir = 'run_%04d' % r
print run_dir
if iteration == None:
iteration = return_last_iteration_integer(
'{r}/output_mnt/'.format(r=run_dir))
f = spimage.sp_image_read(
'{r}/output_mnt/fmodel_{i}.h5'.format(r=run_dir, i=iteration), 0)
if r == r1:
if reference_model == None:
ref = f
else:
ref = reference_model
added_img = f.image[:]
mask = f.mask[:]
if r != r1:
spimage.sp_image_superimpose(ref, f, 0)
added_img += f.image[:]
added_img /= float(len(rundir_range))
new = spimage.sp_image_alloc(*np.shape(added_img))
new.phased = 1
new.mask[:] = mask
new.image[:] = added_img
spimage.sp_image_write(
new, 'avg_fmodel_runs_{i}_{j}_iteration{k}.h5'.format(i=r1,
j=r2,
k=iteration), 0)
new.image[:, :, :] = fft.fftn(added_img)
spimage.sp_image_write(
new, 'avg_model_runs_{i}_{j}_iteration{k}.h5'.format(i=r1,
j=r2,
k=iteration), 0)
def calc_average_img(path, cutoff=None):
if cutoff == None:
run_folders = []
run_folders = select_by_error_cutoff(path, cutoff)['run']
for r in run_folders:
img = spimage.sp_image_read(
r +
'/output_mnt/model_{n}.h5'.format(n=return_last_iteration_integer(
os.path.join(r, 'output_mnt'))), 0)
try:
added_imgs = added_imgs + img.image
except NameError:
added_imgs = img.image
avg_img = added_imgs / float(len(run_folders))
new = spimage.sp_image_alloc(*np.shape(img.image))
new.image[:, :, :] = avg_img
spimage.sp_image_write(
new,
'{p}/average_final_model_{c}_{t}.h5'.format(p=path,
c=cutoff,
t=time.strftime('%Y%m%d')),
0)
def crop_image(in_file, out_file, side, center=None):
"""Function to crop an h5 image and pad with zeros around it"""
img = spimage.sp_image_read(in_file, 0)
if not center:
center = img.detector.image_center[:2]
shifted = 0
if img.shifted:
shifted = 1
img = spimage.sp_image_shift(img)
print "shifted = ", shifted
# cropped = spimage.rectangle_crop(img,lowX,lowY,highX,highY)
cropped = spimage.sp_image_alloc(side, side, 1)
cropped.image[:, :] = image_manipulation.crop_and_pad(img.image, center, side)
cropped.mask[:, :] = image_manipulation.crop_and_pad(img.mask, center, side)
print "did crop"
if shifted:
cropped = spimage.sp_image_shift(cropped)
print "shifted (or not)"
print "write ", out_file
# print "orientation = ", cropped.detector.orientation
# print spimage.sp_3matrix_get(cropped.detector.orientation,0,0,0)
spimage.sp_image_write(cropped, out_file, 16)
spimage.sp_image_free(img)
spimage.sp_image_free(cropped)
print "end"
