def _medianize(dset): num_imgs = get_nr_projs ( dset ) # Return error if there are no images: if (num_imgs == 0): return -1 # Error: # Return the only image in case of one image: elif (num_imgs == 1): return read_tomo(dset,0).astype(float32) # Do the median of all the images if there is more than one image: if num_imgs > 1: # Get first image: im = read_tomo(dset,0).astype(float32) # Read all the remaining files (if any) and save it in a volume: for i in range(1, num_imgs): # Read i-th image from input folder: #im = im + dset[i,:,:].astype(float32) im = im + read_tomo(dset,i).astype(float32) # Medianize volume along the third-dimension: im = im / num_imgs # Reshape output and return: return im.astype(float32)
def _process(lock, int_from, int_to, infile, outfile, outshape, outtype, method, plan, logfilename): # Process the required subset of images: for i in range(int_from, int_to + 1): # Read input image: t0 = time() f_in = getHDF5(infile, 'r') if "/tomo" in f_in: dset = f_in['tomo'] else: dset = f_in['exchange/data'] im = tdf.read_tomo(dset,i).astype(float32) f_in.close() t1 = time() # Perform phase retrieval (first time also PyFFTW prepares a plan): if (method == 0): im = tiehom(im, plan).astype(float32) else: im = phrt(im, plan, method).astype(float32) t2 = time() # Save processed image to HDF5 file (atomic procedure - lock used): _write_data(lock, im, i, outfile, outshape, outtype, logfilename, t2 - t1, t1 - t0)
def _process(lock, int_from, int_to, infile, outfile, outshape, outtype,
method, plan, logfilename):
# Process the required subset of images:
for i in range(int_from, int_to + 1):
# Read input image:
t0 = time()
f_in = getHDF5(infile, 'r')
if "/tomo" in f_in:
dset = f_in['tomo']
else:
dset = f_in['exchange/data']
im = tdf.read_tomo(dset, i).astype(float32)
f_in.close()
t1 = time()
# Perform phase retrieval (first time also PyFFTW prepares a plan):
if (method == 0):
im = tiehom(im, plan).astype(float32)
elif (method == 1):
im = tiehom2020(im, plan).astype(float32)
else:
im = phrt(im, plan, method).astype(float32)
t2 = time()
# Save processed image to HDF5 file (atomic procedure - lock used):
_write_data(lock, im, i, outfile, outshape, outtype, logfilename,
t2 - t1, t1 - t0)
def _process(lock, int_from, int_to, infile, outfile, outshape, outtype, skipflat, plan, flat_end, half_half, half_half_line, dynamic_ff, EFF, filtEFF, im_dark, rotation, interp, border, rescale_factor, logfilename): # Process the required subset of images: for i in range(int_from, int_to + 1): # Read input image: t0 = time() f_in = getHDF5(infile, 'r') if "/tomo" in f_in: dset = f_in['tomo'] else: dset = f_in['exchange/data'] im = tdf.read_tomo(dset,i).astype(float32) f_in.close() t1 = time() # Perform pre-processing (flat fielding, extended FOV, ring removal): if not skipflat: if dynamic_ff: # Dynamic flat fielding with downsampling = 2: im = dynamic_flat_fielding(im, EFF, filtEFF, 2, im_dark) else: im = flat_fielding(im, i, plan, flat_end, half_half, half_half_line, norm_sx, norm_dx) t2 = time() # Rotate: rows, cols = im.shape #if (cols > rows): # angle = - arctan(2.0 / cols) * (rows / 2.0) / 2.0 #else: # angle = - arctan(2.0 / rows) * (cols / 2.0) / 2.0 #print(angle) M = cv2.getRotationMatrix2D((cols / 2, rows / 2), rotation, 1) if interp == 'nearest': interpflag = cv2.INTER_NEAREST elif interp == 'cubic': interpflag = cv2.INTER_CUBIC elif interp == 'lanczos': interpflag = cv2.INTER_LANCZOS4 else: interpflag = cv2.INTER_LINEAR if border == 'constant': borderflag = cv2.BORDER_CONSTANT else: borderflag = cv2.BORDER_REPLICATE im = cv2.warpAffine(im, M, (cols, rows), flags = interpflag, borderMode = borderflag) # Save processed image to HDF5 file (atomic procedure - lock used): _write_data(lock, im, i, outfile, outshape, outtype, rescale_factor, logfilename, t2 - t1, t1 - t0)
def _process(lock, int_from, int_to, infile, dset_str, TIFFFormat, projorder, outpath, outprefix, logfilename): """To do... """ try: f = getHDF5(infile, 'r') dset = f[dset_str] # Process the required subset of images: for i in range(int_from, int_to + 1): # Read input image: t0 = time.time() if projorder: im = tdf.read_tomo(dset, i) else: im = tdf.read_sino(dset, i) # Cast type (if required but it should never occur): if (((im.dtype).type is float64) or ((im.dtype).type is float16)): im = im.astype(float32, copy=False) if (TIFFFormat): fname = outpath + outprefix + '_' + str(i).zfill(4) + '.tif' imsave(fname, im) else: fname = outpath + outprefix + '_' + str(i).zfill(4) + '_' + str(im.shape[1]) + \ 'x' + str(im.shape[0]) + '_' + str(im.dtype) + '.raw' im.tofile(fname) t1 = time.time() # Print out execution time: _write_log(lock, fname, logfilename, t1 - t0) f.close() except Exception: pass
def main(argv):
"""To do...
"""
lock = Lock()
skip_flat = True
first_done = False
pyfftw_cache_disable()
pyfftw_cache_enable()
pyfftw_set_keepalive_time(1800)
# Get the from and to number of files to process:
idx = int(argv[0])
# Get full paths of input TDF and output TDF:
infile = argv[1]
outfile = argv[2]
# Get the phase retrieval parameters:
method = int(argv[3])
param1 = double(argv[4]) # param1( e.g. regParam, or beta)
param2 = double(argv[5]) # param2( e.g. thresh or delta)
energy = double(argv[6])
distance = double(argv[7])
pixsize = double(argv[8]) / 1000.0 # pixsixe from micron to mm:
pad = True if argv[9] == "True" else False
# Tmp path and log file:
tmppath = argv[10]
if not tmppath.endswith(sep): tmppath += sep
logfilename = argv[11]
# Open the HDF5 file and check it contains flat files:
skipflat = False
f_in = getHDF5(infile, 'r')
if "/tomo" in f_in:
dset = f_in['tomo']
if not "/flat" in f_in:
skipflat = True
else:
dset = f_in['exchange/data']
if not "/exchange/data_white" in f_in:
skipflat = True
num_proj = tdf.get_nr_projs(dset)
num_sinos = tdf.get_nr_sinos(dset)
# Check if the HDF5 makes sense:
if (num_proj == 0):
log = open(logfilename, "a")
log.write(linesep + "\tNo projections found. Process will end.")
log.close()
exit()
# Get flats and darks from cache or from file:
if not skipflat:
try:
corrplan = cache2plan(infile, tmppath)
except Exception as e:
#print "Error(s) when reading from cache"
corrplan = extract_flatdark(f_in, True, logfilename)
remove(logfilename)
plan2cache(corrplan, infile, tmppath)
# Read projection:
im = tdf.read_tomo(dset, idx).astype(float32)
f_in.close()
# Apply simple flat fielding (if applicable):
if not skipflat:
if (isinstance(corrplan['im_flat_after'], ndarray)
and isinstance(corrplan['im_flat'], ndarray)
and isinstance(corrplan['im_dark'], ndarray)
and isinstance(corrplan['im_dark_after'], ndarray)):
if (idx < num_proj / 2):
im = (im - corrplan['im_dark']) / (
abs(corrplan['im_flat'] - corrplan['im_dark']) +
finfo(float32).eps)
else:
im = (im - corrplan['im_dark_after']) / (
abs(corrplan['im_flat_after'] - corrplan['im_dark_after'])
+ finfo(float32).eps)
# Prepare plan:
im = im.astype(float32)
if (method == 0):
# Paganin 2002:
plan = tiehom_plan(im, param1, param2, energy, distance, pixsize, pad)
im = tiehom(im, plan).astype(float32)
elif (method == 1):
# Paganin 2020:
plan = tiehom_plan2020(im, param1, param2, energy, distance, pixsize,
pad)
im = tiehom2020(im, plan).astype(float32)
else:
plan = phrt_plan(im, energy, distance, pixsize, param2, param1, method,
pad)
im = phrt(im, plan, method).astype(float32)
# Write down reconstructed preview file (file name modified with metadata):
im = im.astype(float32)
outfile = outfile + '_' + str(im.shape[1]) + 'x' + str(
im.shape[0]) + '_' + str(nanmin(im)) + '$' + str(nanmax(im))
im.tofile(outfile)
def main(argv):
"""To do...
"""
lock = Lock()
skip_flat = True
first_done = False
pyfftw_cache_disable()
pyfftw_cache_enable()
pyfftw_set_keepalive_time(1800)
# Get the from and to number of files to process:
int_from = int(argv[0])
int_to = int(argv[1])
# Get full paths of input TDF and output TDF:
infile = argv[2]
outfile = argv[3]
# Get the phase retrieval parameters:
method = int(argv[4])
param1 = double(argv[5]) # e.g. regParam, or beta
param2 = double(argv[6]) # e.g. thresh or delta
energy = double(argv[7])
distance = double(argv[8])
pixsize = double(argv[9]) / 1000.0 # pixsixe from micron to mm:
pad = True if argv[10] == "True" else False
# Number of threads (actually processes) to use and logfile:
nr_threads = int(argv[11])
logfilename = argv[12]
# Log infos:
log = open(logfilename, "w")
log.write(linesep + "\tInput TDF file: %s" % (infile))
log.write(linesep + "\tOutput TDF file: %s" % (outfile))
log.write(linesep + "\t--------------")
if (method == 0):
log.write(linesep + "\tMethod: TIE-Hom (Paganin et al., 2002)")
log.write(linesep + "\t--------------")
log.write(linesep + "\tDelta/Beta: %0.1f" % ((param2 / param1)))
elif (method == 1):
log.write(linesep +
"\tMethod: Generalized TIE-Hom (Paganin et al., 2020)")
log.write(linesep + "\t--------------")
log.write(linesep + "\tDelta/Beta: %0.1f" % ((param2 / param1)))
else:
log.write(linesep + "\tMethod: Projected CTF (Moosmann et al., 2011)")
log.write(linesep + "\t--------------")
log.write(linesep + "\tRegularization: %0.3f" % (param2))
log.write(linesep + "\tThreshold: %0.3f" % (param1))
log.write(linesep + "\tEnergy: %0.1f keV" % (energy))
log.write(linesep + "\tDistance: %0.1f mm" % (distance))
log.write(linesep + "\tPixel size: %0.3f micron" % (pixsize * 1000))
log.write(linesep + "\t--------------")
log.write(linesep + "\tBrowsing input files...")
log.close()
# Remove a previous copy of output:
if exists(outfile):
remove(outfile)
# Open the HDF5 file:
f_in = getHDF5(infile, 'r')
if "/tomo" in f_in:
dset = f_in['tomo']
else:
dset = f_in['exchange/data']
num_proj = tdf.get_nr_projs(dset)
num_sinos = tdf.get_nr_sinos(dset)
if (num_proj == 0):
log = open(logfilename, "a")
log.write(linesep + "\tNo projections found. Process will end.")
log.close()
exit()
log = open(logfilename, "a")
log.write(linesep + "\tInput files browsed correctly.")
log.close()
# Check extrema (int_to == -1 means all files):
if ((int_to >= num_proj) or (int_to == -1)):
int_to = num_proj - 1
if ((int_from < 0)):
int_from = 0
# Prepare the plan:
log = open(logfilename, "a")
log.write(linesep + "\tPreparing the work plan...")
log.close()
im = tdf.read_tomo(dset, 0).astype(float32)
outshape = tdf.get_dset_shape(im.shape[1], im.shape[0], num_proj)
f_out = getHDF5(outfile, 'w')
f_out_dset = f_out.create_dataset('exchange/data', outshape, float32)
f_out_dset.attrs['min'] = str(amin(im[:]))
f_out_dset.attrs['max'] = str(amax(im[:]))
f_out_dset.attrs['version'] = '1.0'
f_out_dset.attrs['axes'] = "y:theta:x"
f_in.close()
f_out.close()
if (method == 0):
# Paganin 2020:
plan = tiehom_plan(im, param1, param2, energy, distance, pixsize, pad)
elif (method == 1):
# Paganin 2020:
plan = tiehom_plan2020(im, param1, param2, energy, distance, pixsize,
pad)
else:
plan = phrt_plan(im, energy, distance, pixsize, param2, param1, method,
pad)
# Run several threads for independent computation without waiting for threads completion:
for num in range(nr_threads):
start = (num_proj / nr_threads) * num
if (num == nr_threads - 1):
end = num_proj - 1
else:
end = (num_proj / nr_threads) * (num + 1) - 1
Process(target=_process,
args=(lock, start, end, infile, outfile, outshape, float32,
method, plan, logfilename)).start()
def main(argv):
"""To do...
"""
lock = Lock()
skip_flat = True
first_done = False
pyfftw_cache_disable()
pyfftw_cache_enable()
pyfftw_set_keepalive_time(1800)
# Get the from and to number of files to process:
int_from = int(argv[0])
int_to = int(argv[1])
# Get full paths of input TDF and output TDF:
infile = argv[2]
outfile = argv[3]
# Get the phase retrieval parameters:
method = int(argv[4])
param1 = double(argv[5]) # e.g. regParam, or beta
param2 = double(argv[6]) # e.g. thresh or delta
energy = double(argv[7])
distance = double(argv[8])
pixsize = double(argv[9]) / 1000.0 # pixsixe from micron to mm:
pad = True if argv[10] == "True" else False
# Number of threads (actually processes) to use and logfile:
nr_threads = int(argv[11])
logfilename = argv[12]
# Log infos:
log = open(logfilename,"w")
log.write(linesep + "\tInput TDF file: %s" % (infile))
log.write(linesep + "\tOutput TDF file: %s" % (outfile))
log.write(linesep + "\t--------------")
if (method == 0):
log.write(linesep + "\tMethod: TIE-Hom (Paganin et al., 2002)")
log.write(linesep + "\t--------------")
log.write(linesep + "\tDelta/Beta: %0.1f" % ((param2/param1)) )
#else:
# log.write(linesep + "\tMethod: Projected CTF (Moosmann et al., 2011)")
# log.write(linesep + "\t--------------")
# log.write(linesep + "\tDelta/Beta: %0.1f" % ((param2/param1)) )
log.write(linesep + "\tEnergy: %0.1f keV" % (energy))
log.write(linesep + "\tDistance: %0.1f mm" % (distance))
log.write(linesep + "\tPixel size: %0.3f micron" % (pixsize*1000))
log.write(linesep + "\t--------------")
log.write(linesep + "\tBrowsing input files...")
log.close()
# Remove a previous copy of output:
if exists(outfile):
remove(outfile)
# Open the HDF5 file:
f_in = getHDF5(infile, 'r')
if "/tomo" in f_in:
dset = f_in['tomo']
else:
dset = f_in['exchange/data']
num_proj = tdf.get_nr_projs(dset)
num_sinos = tdf.get_nr_sinos(dset)
if (num_proj == 0):
log = open(logfilename,"a")
log.write(linesep + "\tNo projections found. Process will end.")
log.close()
exit()
log = open(logfilename,"a")
log.write(linesep + "\tInput files browsed correctly.")
log.close()
# Check extrema (int_to == -1 means all files):
if ( (int_to >= num_proj) or (int_to == -1) ):
int_to = num_proj - 1
if ( (int_from < 0) ):
int_from = 0
# Prepare the plan:
log = open(logfilename,"a")
log.write(linesep + "\tPreparing the work plan...")
log.close()
im = tdf.read_tomo(dset,0).astype(float32)
outshape = tdf.get_dset_shape(im.shape[1], im.shape[0], num_proj)
f_out = getHDF5(outfile, 'w')
f_out_dset = f_out.create_dataset('exchange/data', outshape, float32)
f_out_dset.attrs['min'] = str(amin(im[:]))
f_out_dset.attrs['max'] = str(amax(im[:]))
f_out_dset.attrs['version'] = '1.0'
f_out_dset.attrs['axes'] = "y:theta:x"
f_in.close()
f_out.close()
if (method == 0):
# Paganin's:
plan = tiehom_plan (im, param1, param2, energy, distance, pixsize, pad)
else:
plan = phrt_plan (im, energy, distance, pixsize, param2, param1, method, pad)
# Run several threads for independent computation without waiting for threads completion:
for num in range(nr_threads):
start = (num_proj / nr_threads)*num
if (num == nr_threads - 1):
end = num_proj - 1
else:
end = (num_proj / nr_threads)*(num + 1) - 1
Process(target=_process, args=(lock, start, end, infile, outfile, outshape, float32, method, plan, logfilename)).start()
def main(argv): """To do... """ lock = Lock() skip_flat = True first_done = False pyfftw_cache_disable() pyfftw_cache_enable() pyfftw_set_keepalive_time(1800) # Get the from and to number of files to process: idx = int(argv[0]) # Get full paths of input TDF and output TDF: infile = argv[1] outfile = argv[2] # Get the phase retrieval parameters: method = int(argv[3]) param1 = double(argv[4]) # param1( e.g. regParam, or beta) param2 = double(argv[5]) # param2( e.g. thresh or delta) energy = double(argv[6]) distance = double(argv[7]) pixsize = double(argv[8]) / 1000.0 # pixsixe from micron to mm: pad = True if argv[9] == "True" else False # Tmp path and log file: tmppath = argv[10] if not tmppath.endswith(sep): tmppath += sep logfilename = argv[11] # Open the HDF5 file and check it contains flat files: skipflat = False f_in = getHDF5(infile, 'r') if "/tomo" in f_in: dset = f_in['tomo'] if not "/flat" in f_in: skipflat = True else: dset = f_in['exchange/data'] if not "/exchange/data_white" in f_in: skipflat = True num_proj = tdf.get_nr_projs(dset) num_sinos = tdf.get_nr_sinos(dset) # Check if the HDF5 makes sense: if (num_proj == 0): log = open(logfilename,"a") log.write(linesep + "\tNo projections found. Process will end.") log.close() exit() # Get flats and darks from cache or from file: if not skipflat: try: corrplan = cache2plan(infile, tmppath) except Exception as e: #print "Error(s) when reading from cache" corrplan = extract_flatdark(f_in, True, logfilename) remove(logfilename) plan2cache(corrplan, infile, tmppath) # Read projection: im = tdf.read_tomo(dset,idx).astype(float32) f_in.close() # Apply simple flat fielding (if applicable): if not skipflat: if (isinstance(corrplan['im_flat_after'], ndarray) and isinstance(corrplan['im_flat'], ndarray) and isinstance(corrplan['im_dark'], ndarray) and isinstance(corrplan['im_dark_after'], ndarray)) : if (idx < num_proj/2): im = (im - corrplan['im_dark']) / (abs(corrplan['im_flat'] - corrplan['im_dark']) + finfo(float32).eps) else: im = (im - corrplan['im_dark_after']) / (abs(corrplan['im_flat_after'] - corrplan['im_dark_after']) + finfo(float32).eps) # Prepare plan: im = im.astype(float32) if (method == 0): # Paganin's: plan = tiehom_plan (im, param1, param2, energy, distance, pixsize, pad) im = tiehom(im, plan).astype(float32) else: plan = phrt_plan (im, energy, distance, pixsize, param2, param1, method, pad) im = phrt(im, plan, method).astype(float32) # Write down reconstructed preview file (file name modified with metadata): im = im.astype(float32) outfile = outfile + '_' + str(im.shape[1]) + 'x' + str(im.shape[0]) + '_' + str( nanmin(im)) + '$' + str( nanmax(im) ) im.tofile(outfile)
def _medianize_withprovenance(dset, provenance_dset, tomoprefix, darkorflatprefix, flagafter): # Get minimum and max timestamps for projection min_t = maxint max_t = -1 for i in range(0, provenance_dset.shape[0]): if provenance_dset["filename", i].startswith(tomoprefix): t = int(mktime(datetime.strptime(provenance_dset["timestamp", i], "%Y-%m-%d %H:%M:%S.%f").timetuple())) if (t < min_t): min_t = t if (t > max_t): max_t = t flag_first = False ct = 0 im = read_tomo(dset,0).astype(float32) for i in range(0, provenance_dset.shape[0]): if provenance_dset["filename", i].startswith(darkorflatprefix): t = int(mktime(datetime.strptime(provenance_dset["timestamp", i], "%Y-%m-%d %H:%M:%S.%f").timetuple())) if flagafter: if (t > max_t): # Get "angle": name = splitext(provenance_dset["filename", i])[0] idx = int(name[-4:]) idx = idx - int(provenance_dset.attrs['first_index']) # Get image and sum it to the series: if not flag_first: flag_first = True im = read_tomo(dset,idx).astype(float32) ct = ct + 1 else: im = im + read_tomo(dset,idx).astype(float32) ct = ct + 1 else: if (t <= min_t): # Get "angle": name = splitext(provenance_dset["filename", i])[0] idx = int(name[-4:]) idx = idx - int(provenance_dset.attrs['first_index']) # Get image and sum it to the series: if not flag_first: flag_first = True im = read_tomo(dset,idx).astype(float32) else: im = im + read_tomo(dset,idx).astype(float32) ct = ct + 1 if ( ct > 0 ): im = im / ct return im.astype(float32) else: return -1 # Error:
def main(argv):
"""Extract a 2D image (projection or sinogram) from the input TDF file (DataExchange HDF5) and
creates a 32-bit RAW file to disk.
Parameters
----------
argv[0] : string
The absolute path of the input TDF.
argv[1] : int
The relative position of the image within the dataset.
argv[2] : string
One of the following options: 'tomo', 'sino', 'flat', 'dark'.
argv[3] : string
The absolute path of the output 32-bit RAW image file. Filename will be modified by adding
image width, image height, minimum and maximum value of the input TDF dataset.
Example
-------
tools_extractdata "S:\\dataset.tdf" 128 tomo "R:\\proj"
"""
try:
#
# Get input parameters:
#
infile = argv[0]
index = int(argv[1])
imtype = argv[2]
outfile = argv[3]
#
# Body
#
# Check if file exists:
if not os.path.exists(infile):
#log = open(logfilename,"a")
#log.write(os.linesep + "\tError: input TDF file not found. Process will end.")
#log.close()
exit()
# Open the HDF5 file:
f = getHDF5(infile, 'r')
if (imtype == 'sino'):
if "/tomo" in f:
dset = f['tomo']
else:
dset = f['exchange/data']
im = tdf.read_sino(dset, index)
elif (imtype == 'dark'):
if "/dark" in f:
dset = f['dark']
else:
dset = f['exchange/data_dark']
im = tdf.read_tomo(dset, index)
elif (imtype == 'flat'):
if "/flat" in f:
dset = f['flat']
else:
dset = f['exchange/data_white']
im = tdf.read_tomo(dset, index)
else:
if "/tomo" in f:
dset = f['tomo']
else:
dset = f['exchange/data']
im = tdf.read_tomo(dset, index)
# Remove Infs e NaNs
tmp = im[:].astype(numpy.float32)
tmp = tmp[numpy.nonzero(numpy.isfinite(tmp))]
# Sort the gray levels:
tmp = numpy.sort(tmp)
# Return as minimum the value the skip 0.30% of "black" tail and 0.05% of "white" tail:
low_idx = int(tmp.shape[0] * 0.0030)
high_idx = int(tmp.shape[0] * 0.9995)
min = tmp[low_idx]
max = tmp[high_idx]
# Modify file name:
outfile = outfile + '_' + str(im.shape[1]) + 'x' + str(
im.shape[0]) + '_' + str(min) + '$' + str(max)
# Cast type:
im = im.astype(float32)
# Write RAW data to disk:
im.tofile(outfile)
except:
exit()
def dff_prepare_plan(white_dset, repetitions, dark):
""" Prepare the Eigen Flat Fields (EFFs) and the filtered EFFs to
be used for dynamic flat fielding.
(Function to be called once before the actual filtering of each projection).
Parameters
----------
white_dset : array_like
3D matrix where each flat field image is stacked along the 3rd dimension.
repetitions: int
Number of iterations to consider during parallel analysis.
dark : array_like
Single dark image (perhaps the average of a series) to be subtracted from
each flat field image. If the images are already dark corrected or dark
correction is not required (e.g. due to a photon counting detector) a matrix
of the proper shape with zeros has to be passed.
Return value
------------
EFF : array_like
Eigen flat fields stacked along the 3rd dimension.
filtEFF : array_like
Filtered eigen flat fields stacked along the 3rd dimension.
Note
----
In this implementation all the collected white field images have to be loaded into
memory and an internal 32-bit copy of the white fields is created. Considering also
that the method better performs with several (i.e. hundreds) flat field images, this
function might raise memory errors.
References
----------
V. Van Nieuwenhove, J. De Beenhouwer, F. De Carlo, L. Mancini, F. Marone,
and J. Sijbers, "Dynamic intensity normalization using eigen flat fields
in X-ray imaging", Optics Express, 23(11), 27975-27989, 2015.
"""
# Get dimensions of flat-field (or white-field) images:
num_flats = get_nr_projs(white_dset) / 4
num_rows = get_nr_sinos(white_dset)
num_cols = get_det_size(white_dset)
# Create local copy of white-field dataset:
tmp_dset = zeros((num_rows * num_cols, num_flats), dtype=float32)
avg = zeros((num_rows * num_cols), dtype=float32)
# For all the flat images:
for i in range(0, tmp_dset.shape[1]):
# Read i-th flat image and dark-correct:
tmp_dset[:, i] = read_tomo(
white_dset,
i).astype(float32).flatten() - dark.astype(float32).flatten()
# Sum the image:
avg = avg + tmp_dset[:, i]
# Compute the mean:
avg = avg / num_flats
# Subtract mean white-field:
for i in range(0, tmp_dset.shape[1]):
tmp_dset[:, i] = tmp_dset[:, i] - avg
# Calculate the number of Eigen Flat Fields (EFF) to use:
V, nrEFF = _parallelAnalysis(tmp_dset, repetitions)
# Compute the EFFs (0-th image is the average "conventional" flat field):
EFF = zeros((num_rows, num_cols, nrEFF + 1), dtype=float32)
EFF[:, :, 0] = avg.reshape((num_rows, num_cols))
for i in range(0, nrEFF):
EFF[:, :, i + 1] = dot(tmp_dset, V[:, num_flats - (i + 1)]).reshape(
(num_rows, num_cols))
# Filter the EFFs:
filtEFF = zeros((num_rows, num_cols, 1 + nrEFF), dtype=float32)
for i in range(1, 1 + nrEFF):
filtEFF[:, :, i] = median_filter(EFF[:, :, i], 3)
return EFF, filtEFF
def main(argv):
"""Try to guess the center of rotation of the input CT dataset.
Parameters
----------
infile : array_like
HDF5 input dataset
outfile : string
Full path where the identified center of rotation will be written as output
scale : int
If sub-pixel precision is interesting, use e.g. 2.0 to get a center of rotation
of .5 value. Use 1.0 if sub-pixel precision is not required
angles : int
Total number of angles of the input dataset
proj_from : int
Initial projections to consider for the assumed angles
proj_to : int
Final projections to consider for the assumed angles
method : string
(not implemented yet)
tmppath : string
Temporary path where look for cached flat/dark files
"""
# Get path:
infile = argv[0] # The HDF5 file on the
outfile = argv[1] # The txt file with the proposed center
scale = float(argv[2])
angles = float(argv[3])
proj_from = int(argv[4])
proj_to = int(argv[5])
method = argv[6]
tmppath = argv[7]
if not tmppath.endswith(sep): tmppath += sep
pyfftw_cache_disable()
pyfftw_cache_enable()
pyfftw_set_keepalive_time(1800)
# Create a silly temporary log:
tmplog = tmppath + basename(infile) + str(time.time())
# Open the HDF5 file (take into account also older TDF versions):
f_in = getHDF5(infile, 'r')
if "/tomo" in f_in:
dset = f_in['tomo']
else:
dset = f_in['exchange/data']
num_proj = tdf.get_nr_projs(dset)
num_sinos = tdf.get_nr_sinos(dset)
# Get flats and darks from cache or from file:
try:
corrplan = cache2plan(infile, tmppath)
except Exception as e:
#print "Error(s) when reading from cache"
corrplan = extract_flatdark(f_in, True, tmplog)
remove(tmplog)
plan2cache(corrplan, infile, tmppath)
# Get first and the 180 deg projections:
im1 = tdf.read_tomo(dset, proj_from).astype(float32)
idx = int(round((proj_to - proj_from) / angles * pi)) + proj_from
im2 = tdf.read_tomo(dset, idx).astype(float32)
# Apply simple flat fielding (if applicable):
if (isinstance(corrplan['im_flat_after'], ndarray)
and isinstance(corrplan['im_flat'], ndarray)
and isinstance(corrplan['im_dark'], ndarray)
and isinstance(corrplan['im_dark_after'], ndarray)):
im1 = ((abs(im1 - corrplan['im_dark'])) /
(abs(corrplan['im_flat'] - corrplan['im_dark']) +
finfo(float32).eps)).astype(float32)
im2 = ((abs(im2 - corrplan['im_dark_after'])) /
(abs(corrplan['im_flat_after'] - corrplan['im_dark_after']) +
finfo(float32).eps)).astype(float32)
# Scale projections (if required) to get subpixel estimation:
if (abs(scale - 1.0) > finfo(float32).eps):
im1 = imresize(im1, (int(round(
scale * im1.shape[0])), int(round(scale * im1.shape[1]))),
interp='bicubic',
mode='F')
im2 = imresize(im2, (int(round(
scale * im2.shape[0])), int(round(scale * im2.shape[1]))),
interp='bicubic',
mode='F')
# Find the center (flipping left-right im2):
cen = findcenter.usecorrelation(im1, im2[:, ::-1])
cen = cen / scale
# Print center to output file:
text_file = open(outfile, "w")
text_file.write(str(int(cen)))
text_file.close()
# Close input HDF5:
f_in.close()
def main(argv):
"""To do...
Usage
-----
Parameters
---------
Example
--------------------------
The following line processes the first ten TIFF files of input path
"/home/in" and saves the processed files to "/home/out" with the
application of the Boin and Haibel filter with smoothing via a Butterworth
filter of order 4 and cutoff frequency 0.01:
destripe /home/in /home/out 1 10 1 0.01 4
"""
lock = Lock()
rescale_factor = 10000.0 # For 16-bit floating point
# Get the from and to number of files to process:
int_from = int(argv[0])
int_to = int(argv[1])
# Get paths:
infile = argv[2]
outfile = argv[3]
# Params for flat fielding with post flats/darks:
flat_end = True if argv[4] == "True" else False
half_half = True if argv[5] == "True" else False
half_half_line = int(argv[6])
# Flat fielding method (conventional or dynamic):
dynamic_ff = True if argv[7] == "True" else False
# Parameters for rotation:
rotation = float(argv[8])
interp = argv[9]
border = argv[10]
# Nr of threads and log file:
nr_threads = int(argv[11])
logfilename = argv[12]
# Log input parameters:
log = open(logfilename,"w")
log.write(linesep + "\tInput TDF file: %s" % (infile))
log.write(linesep + "\tOutput TDF file: %s" % (outfile))
log.write(linesep + "\t--------------")
log.write(linesep + "\tOpening input dataset...")
log.close()
# Remove a previous copy of output:
if exists(outfile):
remove(outfile)
# Open the HDF5 file:
f_in = getHDF5(infile, 'r')
if "/tomo" in f_in:
dset = f_in['tomo']
tomoprefix = 'tomo'
flatprefix = 'flat'
darkprefix = 'dark'
else:
dset = f_in['exchange/data']
if "/provenance/detector_output" in f_in:
prov_dset = f_in['provenance/detector_output']
tomoprefix = prov_dset.attrs['tomo_prefix']
flatprefix = prov_dset.attrs['flat_prefix']
darkprefix = prov_dset.attrs['dark_prefix']
num_proj = tdf.get_nr_projs(dset)
num_sinos = tdf.get_nr_sinos(dset)
if (num_sinos == 0):
log = open(logfilename,"a")
log.write(linesep + "\tNo projections found. Process will end.")
log.close()
exit()
# Check extrema (int_to == -1 means all files):
if ((int_to >= num_proj) or (int_to == -1)):
int_to = num_proj - 1
# Prepare the work plan for flat and dark images:
log = open(logfilename,"a")
log.write(linesep + "\t--------------")
log.write(linesep + "\tPreparing the work plan...")
log.close()
# Extract flat and darks:
skipflat = False
skipdark = False
# Following variables make sense only for dynamic flat fielding:
EFF = -1
filtEFF = -1
im_dark = -1
# Following variable makes sense only for conventional flat fielding:
plan = -1
if not dynamic_ff:
plan = extract_flatdark(f_in, flat_end, logfilename)
if (isscalar(plan['im_flat']) and isscalar(plan['im_flat_after'])):
skipflat = True
else:
skipflat = False
else:
# Dynamic flat fielding:
if "/tomo" in f_in:
if "/flat" in f_in:
flat_dset = f_in['flat']
if "/dark" in f_in:
im_dark = _medianize(f_in['dark'])
else:
skipdark = True
else:
skipflat = True # Nothing to do in this case
else:
if "/exchange/data_white" in f_in:
flat_dset = f_in['/exchange/data_white']
if "/exchange/data_dark" in f_in:
im_dark = _medianize(f_in['/exchange/data_dark'])
else:
skipdark = True
else:
skipflat = True # Nothing to do in this case
# Prepare plan for dynamic flat fielding with 16 repetitions:
if not skipflat:
EFF, filtEFF = dff_prepare_plan(flat_dset, 16, im_dark)
# Get the corrected outshape (in this case it's easy):
im = tdf.read_tomo(dset,0).astype(float32)
outshape = tdf.get_dset_shape(im.shape[1], im.shape[0], num_proj)
# Create the output HDF5 file:
f_out = getHDF5(outfile, 'w')
#f_out_dset = f_out.create_dataset('exchange/data', outshape, im.dtype)
f_out_dset = f_out.create_dataset('exchange/data', outshape, float16)
f_out_dset.attrs['min'] = str(amin(im[:]))
f_out_dset.attrs['max'] = str(amax(im[:]))
f_out_dset.attrs['version'] = '1.0'
f_out_dset.attrs['axes'] = "y:theta:x"
f_out_dset.attrs['rescale_factor'] = str(rescale_factor)
f_out.close()
f_in.close()
# Log infos:
log = open(logfilename,"a")
log.write(linesep + "\tWork plan prepared correctly.")
log.write(linesep + "\t--------------")
log.write(linesep + "\tPerforming pre processing...")
log.close()
# Run several threads for independent computation without waiting for threads
# completion:
for num in range(nr_threads):
start = (num_proj / nr_threads) * num
if (num == nr_threads - 1):
end = num_proj - 1
else:
end = (num_proj / nr_threads) * (num + 1) - 1
Process(target=_process, args=(lock, start, end, infile, outfile, outshape, float16, skipflat, plan,
flat_end, half_half, half_half_line, dynamic_ff, EFF, filtEFF, im_dark, rotation, interp, border,
rescale_factor, logfilename)).start()
def main(argv):
"""To do...
Usage
-----
Parameters
---------
Example
--------------------------
The following line processes the first ten TIFF files of input path
"/home/in" and saves the processed files to "/home/out" with the
application of the Boin and Haibel filter with smoothing via a Butterworth
filter of order 4 and cutoff frequency 0.01:
destripe /home/in /home/out 1 10 1 0.01 4
"""
lock = Lock()
rescale_factor = 10000.0 # For 16-bit floating point
# Get the from and to number of files to process:
int_from = int(argv[0])
int_to = int(argv[1])
# Get paths:
infile = argv[2]
outfile = argv[3]
# Params for flat fielding with post flats/darks:
flat_end = True if argv[4] == "True" else False
half_half = True if argv[5] == "True" else False
half_half_line = int(argv[6])
# Flat fielding method (conventional or dynamic):
dynamic_ff = True if argv[7] == "True" else False
# Parameters for rotation:
rotation = float(argv[8])
interp = argv[9]
border = argv[10]
# Nr of threads and log file:
nr_threads = int(argv[11])
logfilename = argv[12]
# Log input parameters:
log = open(logfilename, "w")
log.write(linesep + "\tInput TDF file: %s" % (infile))
log.write(linesep + "\tOutput TDF file: %s" % (outfile))
log.write(linesep + "\t--------------")
log.write(linesep + "\tOpening input dataset...")
log.close()
# Remove a previous copy of output:
if exists(outfile):
remove(outfile)
# Open the HDF5 file:
f_in = getHDF5(infile, 'r')
if "/tomo" in f_in:
dset = f_in['tomo']
tomoprefix = 'tomo'
flatprefix = 'flat'
darkprefix = 'dark'
else:
dset = f_in['exchange/data']
if "/provenance/detector_output" in f_in:
prov_dset = f_in['provenance/detector_output']
tomoprefix = prov_dset.attrs['tomo_prefix']
flatprefix = prov_dset.attrs['flat_prefix']
darkprefix = prov_dset.attrs['dark_prefix']
num_proj = tdf.get_nr_projs(dset)
num_sinos = tdf.get_nr_sinos(dset)
if (num_sinos == 0):
log = open(logfilename, "a")
log.write(linesep + "\tNo projections found. Process will end.")
log.close()
exit()
# Check extrema (int_to == -1 means all files):
if ((int_to >= num_proj) or (int_to == -1)):
int_to = num_proj - 1
# Prepare the work plan for flat and dark images:
log = open(logfilename, "a")
log.write(linesep + "\t--------------")
log.write(linesep + "\tPreparing the work plan...")
log.close()
# Extract flat and darks:
skipflat = False
skipdark = False
# Following variables make sense only for dynamic flat fielding:
EFF = -1
filtEFF = -1
im_dark = -1
# Following variable makes sense only for conventional flat fielding:
plan = -1
if not dynamic_ff:
plan = extract_flatdark(f_in, flat_end, logfilename)
if (isscalar(plan['im_flat']) and isscalar(plan['im_flat_after'])):
skipflat = True
else:
skipflat = False
else:
# Dynamic flat fielding:
if "/tomo" in f_in:
if "/flat" in f_in:
flat_dset = f_in['flat']
if "/dark" in f_in:
im_dark = _medianize(f_in['dark'])
else:
skipdark = True
else:
skipflat = True # Nothing to do in this case
else:
if "/exchange/data_white" in f_in:
flat_dset = f_in['/exchange/data_white']
if "/exchange/data_dark" in f_in:
im_dark = _medianize(f_in['/exchange/data_dark'])
else:
skipdark = True
else:
skipflat = True # Nothing to do in this case
# Prepare plan for dynamic flat fielding with 16 repetitions:
if not skipflat:
EFF, filtEFF = dff_prepare_plan(flat_dset, 16, im_dark)
# Get the corrected outshape (in this case it's easy):
im = tdf.read_tomo(dset, 0).astype(float32)
outshape = tdf.get_dset_shape(im.shape[1], im.shape[0], num_proj)
# Create the output HDF5 file:
f_out = getHDF5(outfile, 'w')
#f_out_dset = f_out.create_dataset('exchange/data', outshape, im.dtype)
f_out_dset = f_out.create_dataset('exchange/data', outshape, float16)
f_out_dset.attrs['min'] = str(amin(im[:]))
f_out_dset.attrs['max'] = str(amax(im[:]))
f_out_dset.attrs['version'] = '1.0'
f_out_dset.attrs['axes'] = "y:theta:x"
f_out_dset.attrs['rescale_factor'] = str(rescale_factor)
f_out.close()
f_in.close()
# Log infos:
log = open(logfilename, "a")
log.write(linesep + "\tWork plan prepared correctly.")
log.write(linesep + "\t--------------")
log.write(linesep + "\tPerforming pre processing...")
log.close()
# Run several threads for independent computation without waiting for threads
# completion:
for num in range(nr_threads):
start = (num_proj / nr_threads) * num
if (num == nr_threads - 1):
end = num_proj - 1
else:
end = (num_proj / nr_threads) * (num + 1) - 1
Process(target=_process,
args=(lock, start, end, infile, outfile, outshape, float16,
skipflat, plan, flat_end, half_half, half_half_line,
dynamic_ff, EFF, filtEFF, im_dark, rotation, interp,
border, rescale_factor, logfilename)).start()
def main(argv):
"""Try to guess the amount of overlap in the case of extended FOV CT.
Parameters
----------
infile : array_like
HDF5 input dataset
outfile : string
Full path where the identified overlap will be written as output
scale : int
If sub-pixel precision is interesting, use e.g. 2.0 to get an overlap
of .5 value. Use 1.0 if sub-pixel precision is not required
tmppath : int
Temporary path where look for cached flat/dark files
"""
# Get path:
infile = argv[0] # The HDF5 file on the SSD
outfile = argv[1] # The txt file with the proposed center
scale = float(argv[2])
tmppath = argv[3]
if not tmppath.endswith(sep): tmppath += sep
# Create a silly temporary log:
tmplog = tmppath + basename(infile) + str(time.time())
# Open the HDF5 file:
f_in = getHDF5(infile, 'r')
if "/tomo" in f_in:
dset = f_in['tomo']
else:
dset = f_in['exchange/data']
num_proj = tdf.get_nr_projs(dset)
# Get first and 180 deg projections:
im1 = tdf.read_tomo(dset, 0).astype(float32)
im2 = tdf.read_tomo(dset, num_proj / 2).astype(float32)
# Get flats and darks from cache or from file:
try:
corrplan = cache2plan(infile, tmppath)
except Exception as e:
#print "Error(s) when reading from cache"
corrplan = extract_flatdark(f_in, True, tmplog)
remove(tmplog)
plan2cache(corrplan, infile, tmppath)
# Apply simple flat fielding (if applicable):
if (isinstance(corrplan['im_flat_after'], ndarray)
and isinstance(corrplan['im_flat'], ndarray)
and isinstance(corrplan['im_dark'], ndarray)
and isinstance(corrplan['im_dark_after'], ndarray)):
im1 = ((abs(im1 - corrplan['im_dark'])) /
(abs(corrplan['im_flat'] - corrplan['im_dark']) +
finfo(float32).eps)).astype(float32)
im2 = ((abs(im2 - corrplan['im_dark_after'])) /
(abs(corrplan['im_flat_after'] - corrplan['im_dark_after']) +
finfo(float32).eps)).astype(float32)
# Scale projections (if required) to get subpixel estimation:
if (abs(scale - 1.0) > finfo(float32).eps):
im1 = imresize(im1, (int(round(
scale * im1.shape[0])), int(round(scale * im1.shape[1]))),
interp='bicubic',
mode='F')
im2 = imresize(im2, (int(round(
scale * im2.shape[0])), int(round(scale * im2.shape[1]))),
interp='bicubic',
mode='F')
# Find the center (flipping left-right im2): DISTINGUISH BETWEEN AIR ON THE RIGHT AND ON THE LEFT??????
cen = findcenter.usecorrelation(im1, im2[:, ::-1])
cen = (cen / scale) * 2.0
# Print center to output file:
text_file = open(outfile, "w")
text_file.write(str(int(abs(cen))))
text_file.close()
# Close input HDF5:
f_in.close()
def dff_prepare_plan(white_dset, repetitions, dark): """ Prepare the Eigen Flat Fields (EFFs) and the filtered EFFs to be used for dynamic flat fielding. (Function to be called once before the actual filtering of each projection). Parameters ---------- white_dset : array_like 3D matrix where each flat field image is stacked along the 3rd dimension. repetitions: int Number of iterations to consider during parallel analysis. dark : array_like Single dark image (perhaps the average of a series) to be subtracted from each flat field image. If the images are already dark corrected or dark correction is not required (e.g. due to a photon counting detector) a matrix of the proper shape with zeros has to be passed. Return value ------------ EFF : array_like Eigen flat fields stacked along the 3rd dimension. filtEFF : array_like Filtered eigen flat fields stacked along the 3rd dimension. Note ---- In this implementation all the collected white field images have to be loaded into memory and an internal 32-bit copy of the white fields is created. Considering also that the method better performs with several (i.e. hundreds) flat field images, this function might raise memory errors. References ---------- V. Van Nieuwenhove, J. De Beenhouwer, F. De Carlo, L. Mancini, F. Marone, and J. Sijbers, "Dynamic intensity normalization using eigen flat fields in X-ray imaging", Optics Express, 23(11), 27975-27989, 2015. """ # Get dimensions of flat-field (or white-field) images: num_flats = get_nr_projs(white_dset)/4 num_rows = get_nr_sinos(white_dset) num_cols = get_det_size(white_dset) # Create local copy of white-field dataset: tmp_dset = zeros((num_rows * num_cols, num_flats), dtype=float32) avg = zeros((num_rows * num_cols), dtype=float32) # For all the flat images: for i in range(0, tmp_dset.shape[1]): # Read i-th flat image and dark-correct: tmp_dset[:,i] = read_tomo(white_dset,i).astype(float32).flatten() - dark.astype(float32).flatten() # Sum the image: avg = avg + tmp_dset[:,i] # Compute the mean: avg = avg / num_flats # Subtract mean white-field: for i in range(0, tmp_dset.shape[1]): tmp_dset[:,i] = tmp_dset[:,i] - avg # Calculate the number of Eigen Flat Fields (EFF) to use: V, nrEFF = _parallelAnalysis(tmp_dset, repetitions) # Compute the EFFs (0-th image is the average "conventional" flat field): EFF = zeros((num_rows, num_cols, nrEFF + 1), dtype=float32) EFF[:,:,0] = avg.reshape((num_rows, num_cols)) for i in range(0, nrEFF): EFF[:,:,i + 1] = dot(tmp_dset, V[:,num_flats - (i + 1)]).reshape((num_rows, num_cols)) # Filter the EFFs: filtEFF = zeros((num_rows, num_cols, 1 + nrEFF), dtype=float32) for i in range(1, 1 + nrEFF): filtEFF[:,:,i] = median_filter(EFF[:,:,i], 3) return EFF, filtEFF
def main(argv):
"""To do...
"""
# Get the zero-order index of the sinogram to pre-process:
idx = int(argv[0])
# Get paths:
infile = argv[1]
outfile = argv[2]
# Normalization parameters:
norm_sx = int(argv[3])
norm_dx = int(argv[4])
# Params for flat fielding with post flats/darks:
flat_end = True if argv[5] == "True" else False
half_half = True if argv[6] == "True" else False
half_half_line = int(argv[7])
# Flat fielding method (conventional or dynamic):
dynamic_ff = True if argv[8] == "True" else False
# Parameters for rotation:
rotation = float(argv[9])
interp = argv[10]
border = argv[11]
# Tmp path and log file:
tmppath = argv[12]
if not tmppath.endswith(sep): tmppath += sep
logfilename = argv[13]
# Open the HDF5 file:
f_in = getHDF5(infile, 'r')
try:
if "/tomo" in f_in:
dset = f_in['tomo']
else:
dset = f_in['exchange/data']
except:
log = open(logfilename, "a")
log.write(linesep + "\tError reading input dataset. Process will end.")
log.close()
exit()
num_proj = tdf.get_nr_projs(dset)
num_sinos = tdf.get_nr_sinos(dset)
# Check if the HDF5 makes sense:
if (num_sinos == 0):
log = open(logfilename, "a")
log.write(linesep + "\tNo projections found. Process will end.")
log.close()
exit()
# Get flat and darks from cache or from file:
skipflat = False
skipdark = False
if not dynamic_ff:
try:
corrplan = cache2plan(infile, tmppath)
except Exception as e:
#print "Error(s) when reading from cache"
corrplan = extract_flatdark(f_in, flat_end, logfilename)
if (isscalar(corrplan['im_flat'])
and isscalar(corrplan['im_flat_after'])):
skipflat = True
else:
plan2cache(corrplan, infile, tmppath)
else:
# Dynamic flat fielding:
if "/tomo" in f_in:
if "/flat" in f_in:
flat_dset = f_in['flat']
if "/dark" in f_in:
im_dark = _medianize(f_in['dark'])
else:
skipdark = True
else:
skipflat = True # Nothing to do in this case
else:
if "/exchange/data_white" in f_in:
flat_dset = f_in['/exchange/data_white']
if "/exchange/data_dark" in f_in:
im_dark = _medianize(f_in['/exchange/data_dark'])
else:
skipdark = True
else:
skipflat = True # Nothing to do in this case
# Prepare plan for dynamic flat fielding with 16 repetitions:
if not skipflat:
EFF, filtEFF = dff_prepare_plan(flat_dset, 16, im_dark)
# Read input image:
im = tdf.read_tomo(dset, idx).astype(float32)
f_in.close()
# Perform pre-processing (flat fielding, extended FOV, ring removal):
if not skipflat:
if dynamic_ff:
# Dynamic flat fielding with downsampling = 2:
im = dynamic_flat_fielding(im, EFF, filtEFF, 2, im_dark)
else:
im = flat_fielding(im, idx, corrplan, flat_end, half_half,
half_half_line, norm_sx, norm_dx)
# Rotate:
rows, cols = im.shape
M = cv2.getRotationMatrix2D((cols / 2, rows / 2), rotation, 1)
if interp == 'nearest':
interpflag = cv2.INTER_NEAREST
elif interp == 'cubic':
interpflag = cv2.INTER_CUBIC
elif interp == 'lanczos':
interpflag = cv2.INTER_LANCZOS4
else:
interpflag = cv2.INTER_LINEAR
if border == 'constant':
borderflag = cv2.BORDER_CONSTANT
else:
borderflag = cv2.BORDER_REPLICATE
im = cv2.warpAffine(im,
M, (cols, rows),
flags=interpflag,
borderMode=borderflag)
# Write down reconstructed preview file (file name modified with metadata):
im = im.astype(float32)
outfile2 = outfile + '_' + str(im.shape[1]) + 'x' + str(
im.shape[0]) + '_' + str(nanmin(im)) + '$' + str(
nanmax(im)) + '_after.raw'
im.tofile(outfile2)
def main(argv):
"""To do...
Usage
-----
Parameters
---------
Example
--------------------------
"""
lock = Lock()
# Get the from and to number of files to process:
int_from = int(argv[0])
int_to = int(argv[1])
# Get paths:
infile_1 = argv[2]
infile_2 = argv[3]
infile_3 = argv[4]
outfile_abs = argv[5]
outfile_ref = argv[6]
outfile_sca = argv[7]
# Normalization parameters:
norm_sx = int(argv[8])
norm_dx = int(argv[9])
# Params for flat fielding with post flats/darks:
flat_end = True if argv[10] == "True" else False
half_half = True if argv[11] == "True" else False
half_half_line = int(argv[12])
# Params for extended FOV:
ext_fov = True if argv[13] == "True" else False
ext_fov_rot_right = argv[14]
if ext_fov_rot_right == "True":
ext_fov_rot_right = True
if (ext_fov):
norm_sx = 0
else:
ext_fov_rot_right = False
if (ext_fov):
norm_dx = 0
ext_fov_overlap = int(argv[15])
ext_fov_normalize = True if argv[16] == "True" else False
ext_fov_average = True if argv[17] == "True" else False
# Method and parameters coded into a string:
ringrem = argv[18]
# Flat fielding method (conventional or dynamic):
dynamic_ff = True if argv[19] == "True" else False
# Shift parameters:
shiftVert_1 = int(argv[20])
shiftHoriz_1 = int(argv[21])
shiftVert_2 = int(argv[22])
shiftHoriz_2 = int(argv[23])
shiftVert_3 = int(argv[24])
shiftHoriz_3 = int(argv[25])
# DEI coefficients:
r1 = float(argv[26])
r2 = float(argv[27])
r3 = float(argv[28])
d1 = float(argv[29])
d2 = float(argv[30])
d3 = float(argv[31])
dd1 = float(argv[32])
dd2 = float(argv[33])
dd3 = float(argv[34])
# Nr of threads and log file:
nr_threads = int(argv[35])
logfilename = argv[36]
# Log input parameters:
log = open(logfilename, "w")
log.write(linesep + "\tInput TDF file #1: %s" % (infile_1))
log.write(linesep + "\tInput TDF file #2: %s" % (infile_2))
log.write(linesep + "\tInput TDF file #3: %s" % (infile_3))
log.write(linesep + "\tOutput TDF file for Absorption: %s" % (outfile_abs))
log.write(linesep + "\tOutput TDF file for Refraction: %s" % (outfile_ref))
log.write(linesep + "\tOutput TDF file for Scattering: %s" % (outfile_sca))
log.write(linesep + "\t--------------")
log.write(linesep + "\tOpening input dataset...")
log.close()
# Remove a previous copy of output:
#if exists(outfile):
# remove(outfile)
# Open the HDF5 files:
f_in_1 = getHDF5(infile_1, 'r')
f_in_2 = getHDF5(infile_2, 'r')
f_in_3 = getHDF5(infile_3, 'r')
if "/tomo" in f_in_1:
dset_1 = f_in_1['tomo']
tomoprefix_1 = 'tomo'
flatprefix_1 = 'flat'
darkprefix_1 = 'dark'
else:
dset_1 = f_in_1['exchange/data']
if "/provenance/detector_output" in f_in_1:
prov_dset_1 = f_in_1['provenance/detector_output']
tomoprefix_1 = prov_dset_1.attrs['tomo_prefix']
flatprefix_1 = prov_dset_1.attrs['flat_prefix']
darkprefix_1 = prov_dset_1.attrs['dark_prefix']
if "/tomo" in f_in_2:
dset_2 = f_in_2['tomo']
tomoprefix_2 = 'tomo'
flatprefix_2 = 'flat'
darkprefix_2 = 'dark'
else:
dset_2 = f_in_2['exchange/data']
if "/provenance/detector_output" in f_in_2:
prov_dset_2 = f_in_2['provenance/detector_output']
tomoprefix_2 = prov_dset_2.attrs['tomo_prefix']
flatprefix_2 = prov_dset_2.attrs['flat_prefix']
darkprefix_2 = prov_dset_2.attrs['dark_prefix']
if "/tomo" in f_in_3:
dset_3 = f_in_3['tomo']
tomoprefix_3 = 'tomo'
flatprefix_3 = 'flat'
darkprefix_3 = 'dark'
else:
dset_3 = f_in_3['exchange/data']
if "/provenance/detector_output" in f_in_3:
prov_dset_3 = f_in_1['provenance/detector_output']
tomoprefix_3 = prov_dset_3.attrs['tomo_prefix']
flatprefix_3 = prov_dset_3.attrs['flat_prefix']
darkprefix_3 = prov_dset_3.attrs['dark_prefix']
# Assuming that what works for the dataset #1 works for the other two:
num_proj = tdf.get_nr_projs(dset_1)
num_sinos = tdf.get_nr_sinos(dset_1)
if (num_sinos == 0):
log = open(logfilename, "a")
log.write(linesep + "\tNo projections found. Process will end.")
log.close()
exit()
# Check extrema (int_to == -1 means all files):
if ((int_to >= num_sinos) or (int_to == -1)):
int_to = num_sinos - 1
# Prepare the work plan for flat and dark images:
log = open(logfilename, "a")
log.write(linesep + "\t--------------")
log.write(linesep + "\tPreparing the work plan...")
log.close()
# Extract flat and darks:
skipflat_1 = False
skipdark_1 = False
skipflat_2 = False
skipdark_2 = False
skipflat_3 = False
skipdark_3 = False
# Following variables make sense only for dynamic flat fielding:
EFF_1 = -1
filtEFF_1 = -1
im_dark_1 = -1
EFF_2 = -1
filtEFF_2 = -1
im_dark_2 = -1
EFF_3 = -1
filtEFF_3 = -1
im_dark_3 = -1
# Following variable makes sense only for conventional flat fielding:
plan_1 = -1
plan_2 = -1
plan_3 = -1
if not dynamic_ff:
plan_1 = extract_flatdark(f_in_1, flat_end, logfilename)
if (isscalar(plan_1['im_flat']) and isscalar(plan_1['im_flat_after'])):
skipflat_1 = True
else:
skipflat_1 = False
plan_2 = extract_flatdark(f_in_2, flat_end, logfilename)
if (isscalar(plan_2['im_flat']) and isscalar(plan_2['im_flat_after'])):
skipflat_2 = True
else:
skipflat_2 = False
plan_3 = extract_flatdark(f_in_3, flat_end, logfilename)
if (isscalar(plan_3['im_flat']) and isscalar(plan_3['im_flat_after'])):
skipflat_3 = True
else:
skipflat_3 = False
else:
# Dynamic flat fielding:
if "/tomo" in f_in_1:
if "/flat" in f_in_1:
flat_dset_1 = f_in_1['flat']
if "/dark" in f_in_1:
im_dark_1 = _medianize(f_in_1['dark'])
else:
skipdark_1 = True
else:
skipflat_1 = True # Nothing to do in this case
else:
if "/exchange/data_white" in f_in_1:
flat_dset_1 = f_in_1['/exchange/data_white']
if "/exchange/data_dark" in f_in_1:
im_dark_1 = _medianize(f_in_1['/exchange/data_dark'])
else:
skipdark_1 = True
else:
skipflat_1 = True # Nothing to do in this case
# Prepare plan for dynamic flat fielding with 16 repetitions:
if not skipflat_1:
EFF_1, filtEFF_1 = dff_prepare_plan(flat_dset_1, 16, im_dark_1)
# Dynamic flat fielding:
if "/tomo" in f_in_2:
if "/flat" in f_in_2:
flat_dset_2 = f_in_2['flat']
if "/dark" in f_in_2:
im_dark_2 = _medianize(f_in_2['dark'])
else:
skipdark_2 = True
else:
skipflat_2 = True # Nothing to do in this case
else:
if "/exchange/data_white" in f_in_2:
flat_dset_2 = f_in_2['/exchange/data_white']
if "/exchange/data_dark" in f_in_2:
im_dark_2 = _medianize(f_in_2['/exchange/data_dark'])
else:
skipdark_2 = True
else:
skipflat_2 = True # Nothing to do in this case
# Prepare plan for dynamic flat fielding with 16 repetitions:
if not skipflat_2:
EFF_2, filtEFF_2 = dff_prepare_plan(flat_dset_2, 16, im_dark_2)
# Dynamic flat fielding:
if "/tomo" in f_in_3:
if "/flat" in f_in_3:
flat_dset_3 = f_in_3['flat']
if "/dark" in f_in_3:
im_dark_3 = _medianize(f_in_3['dark'])
else:
skipdark_3 = True
else:
skipflat_3 = True # Nothing to do in this case
else:
if "/exchange/data_white" in f_in_3:
flat_dset_3 = f_in_3['/exchange/data_white']
if "/exchange/data_dark" in f_in_3:
im_dark_3 = _medianize(f_in_3['/exchange/data_dark'])
else:
skipdark_3 = True
else:
skipflat_3 = True # Nothing to do in this case
# Prepare plan for dynamic flat fielding with 16 repetitions:
if not skipflat_3:
EFF_3, filtEFF_3 = dff_prepare_plan(flat_dset_3, 16, im_dark_3)
# Outfile shape can be determined only after first processing in ext FOV mode:
if (ext_fov):
# Read input sino:
idx = num_sinos / 2
im = tdf.read_sino(dset_1, idx).astype(float32)
im = extfov_correction(im, ext_fov_rot_right, ext_fov_overlap,
ext_fov_normalize, ext_fov_average)
# Get the corrected outshape:
outshape = tdf.get_dset_shape(im.shape[1], num_sinos, im.shape[0])
else:
# Get the corrected outshape (in this case it's easy):
im = tdf.read_tomo(dset_1, 0).astype(float32)
outshape = tdf.get_dset_shape(im.shape[1], im.shape[0], num_proj)
f_in_1.close()
f_in_2.close()
f_in_3.close()
# Create the output HDF5 files:
f_out_abs = getHDF5(outfile_abs, 'w')
f_out_dset_abs = f_out_abs.create_dataset('exchange/data', outshape,
float32)
f_out_dset_abs.attrs['min'] = str(finfo(float32).max)
f_out_dset_abs.attrs['max'] = str(finfo(float32).min)
f_out_dset_abs.attrs['version'] = '1.0'
f_out_dset_abs.attrs['axes'] = "y:theta:x"
f_out_abs.close()
f_out_ref = getHDF5(outfile_ref, 'w')
f_out_dset_ref = f_out_ref.create_dataset('exchange/data', outshape,
float32)
f_out_dset_ref.attrs['min'] = str(finfo(float32).max)
f_out_dset_ref.attrs['max'] = str(finfo(float32).min)
f_out_dset_ref.attrs['version'] = '1.0'
f_out_dset_ref.attrs['axes'] = "y:theta:x"
f_out_ref.close()
f_out_sca = getHDF5(outfile_sca, 'w')
f_out_dset_sca = f_out_sca.create_dataset('exchange/data', outshape,
float32)
f_out_dset_sca.attrs['min'] = str(finfo(float32).max)
f_out_dset_sca.attrs['max'] = str(finfo(float32).min)
f_out_dset_sca.attrs['version'] = '1.0'
f_out_dset_sca.attrs['axes'] = "y:theta:x"
f_out_sca.close()
# Log infos:
log = open(logfilename, "a")
log.write(linesep + "\tWork plan prepared correctly.")
log.write(linesep + "\t--------------")
log.write(linesep + "\tPerforming GDEI...")
log.close()
# Run several threads for independent computation without waiting for threads
# completion:
for num in range(nr_threads):
start = (num_sinos / nr_threads) * num
if (num == nr_threads - 1):
end = num_sinos - 1
else:
end = (num_sinos / nr_threads) * (num + 1) - 1
Process(
target=_process,
args=(lock, start, end, num_sinos, infile_1, infile_2, infile_3,
outfile_abs, outfile_ref, outfile_sca, r1, r2, r3, d1, d2,
d3, dd1, dd2, dd3, shiftVert_1, shiftHoriz_1, shiftVert_2,
shiftHoriz_2, shiftVert_3, shiftHoriz_3, outshape, float32,
skipflat_1, skipflat_2, skipflat_3, plan_1, plan_2, plan_3,
norm_sx, norm_dx, flat_end, half_half, half_half_line,
ext_fov, ext_fov_rot_right, ext_fov_overlap,
ext_fov_normalize, ext_fov_average, ringrem, dynamic_ff,
EFF_1, EFF_2, EFF_3, filtEFF_1, filtEFF_2, filtEFF_3,
im_dark_1, im_dark_2, im_dark_3, logfilename)).start()
def _process(lock, int_from, int_to, infile, outfile, outshape, outtype,
skipflat, plan, flat_end, half_half, half_half_line, dynamic_ff,
EFF, filtEFF, im_dark, rotation, interp, border, rescale_factor,
logfilename):
# Process the required subset of images:
for i in range(int_from, int_to + 1):
# Read input image:
t0 = time()
f_in = getHDF5(infile, 'r')
if "/tomo" in f_in:
dset = f_in['tomo']
else:
dset = f_in['exchange/data']
im = tdf.read_tomo(dset, i).astype(float32)
f_in.close()
t1 = time()
# Perform pre-processing (flat fielding, extended FOV, ring removal):
if not skipflat:
if dynamic_ff:
# Dynamic flat fielding with downsampling = 2:
im = dynamic_flat_fielding(im, EFF, filtEFF, 2, im_dark)
else:
im = flat_fielding(im, i, plan, flat_end, half_half,
half_half_line, norm_sx, norm_dx)
t2 = time()
# Rotate:
rows, cols = im.shape
#if (cols > rows):
# angle = - arctan(2.0 / cols) * (rows / 2.0) / 2.0
#else:
# angle = - arctan(2.0 / rows) * (cols / 2.0) / 2.0
#print(angle)
M = cv2.getRotationMatrix2D((cols / 2, rows / 2), rotation, 1)
if interp == 'nearest':
interpflag = cv2.INTER_NEAREST
elif interp == 'cubic':
interpflag = cv2.INTER_CUBIC
elif interp == 'lanczos':
interpflag = cv2.INTER_LANCZOS4
else:
interpflag = cv2.INTER_LINEAR
if border == 'constant':
borderflag = cv2.BORDER_CONSTANT
else:
borderflag = cv2.BORDER_REPLICATE
im = cv2.warpAffine(im,
M, (cols, rows),
flags=interpflag,
borderMode=borderflag)
# Save processed image to HDF5 file (atomic procedure - lock used):
_write_data(lock, im, i, outfile, outshape, outtype, rescale_factor,
logfilename, t2 - t1, t1 - t0)
def main(argv):
"""Try to guess the amount of overlap in the case of extended FOV CT.
Parameters
----------
infile : array_like
HDF5 input dataset
outfile : string
Full path where the identified overlap will be written as output
scale : int
If sub-pixel precision is interesting, use e.g. 2.0 to get an overlap
of .5 value. Use 1.0 if sub-pixel precision is not required
tmppath : int
Temporary path where look for cached flat/dark files
"""
# Get path:
infile = argv[0] # The HDF5 file on the SSD
outfile = argv[1] # The txt file with the proposed center
scale = float(argv[2])
tmppath = argv[3]
if not tmppath.endswith(sep): tmppath += sep
# Create a silly temporary log:
tmplog = tmppath + basename(infile) + str(time.time())
# Open the HDF5 file:
f_in = getHDF5( infile, 'r' )
if "/tomo" in f_in:
dset = f_in['tomo']
else:
dset = f_in['exchange/data']
num_proj = tdf.get_nr_projs(dset)
# Get first and 180 deg projections:
im1 = tdf.read_tomo(dset,0).astype(float32)
im2 = tdf.read_tomo(dset,num_proj/2).astype(float32)
# Get flats and darks from cache or from file:
try:
corrplan = cache2plan(infile, tmppath)
except Exception as e:
#print "Error(s) when reading from cache"
corrplan = extract_flatdark(f_in, True, tmplog)
remove(tmplog)
plan2cache(corrplan, infile, tmppath)
# Apply simple flat fielding (if applicable):
if (isinstance(corrplan['im_flat_after'], ndarray) and isinstance(corrplan['im_flat'], ndarray) and
isinstance(corrplan['im_dark'], ndarray) and isinstance(corrplan['im_dark_after'], ndarray)) :
im1 = ((abs(im1 - corrplan['im_dark'])) / (abs(corrplan['im_flat'] - corrplan['im_dark']) + finfo(float32).eps)).astype(float32)
im2 = ((abs(im2 - corrplan['im_dark_after'])) / (abs(corrplan['im_flat_after'] - corrplan['im_dark_after']) + finfo(float32).eps)).astype(float32)
# Scale projections (if required) to get subpixel estimation:
if ( abs(scale - 1.0) > finfo(float32).eps ):
im1 = imresize(im1, (int(round(scale*im1.shape[0])), int(round(scale*im1.shape[1]))), interp='bicubic', mode='F');
im2 = imresize(im2, (int(round(scale*im2.shape[0])), int(round(scale*im2.shape[1]))), interp='bicubic', mode='F');
# Find the center (flipping left-right im2): DISTINGUISH BETWEEN AIR ON THE RIGHT AND ON THE LEFT??????
cen = findcenter.usecorrelation(im1, im2[ :,::-1])
cen = (cen / scale)*2.0
# Print center to output file:
text_file = open(outfile, "w")
text_file.write(str(int(abs(cen))))
text_file.close()
# Close input HDF5:
f_in.close()
def main(argv):
"""Try to guess the center of rotation of the input CT dataset.
Parameters
----------
infile : array_like
HDF5 input dataset
outfile : string
Full path where the identified center of rotation will be written as output
scale : int
If sub-pixel precision is interesting, use e.g. 2.0 to get a center of rotation
of .5 value. Use 1.0 if sub-pixel precision is not required
angles : int
Total number of angles of the input dataset
proj_from : int
Initial projections to consider for the assumed angles
proj_to : int
Final projections to consider for the assumed angles
method : string
(not implemented yet)
tmppath : string
Temporary path where look for cached flat/dark files
"""
# Get path:
infile = argv[0] # The HDF5 file on the
outfile = argv[1] # The txt file with the proposed center
scale = float(argv[2])
angles = float(argv[3])
proj_from = int(argv[4])
proj_to = int(argv[5])
method = argv[6]
tmppath = argv[7]
if not tmppath.endswith(sep): tmppath += sep
pyfftw_cache_disable()
pyfftw_cache_enable()
pyfftw_set_keepalive_time(1800)
# Create a silly temporary log:
tmplog = tmppath + basename(infile) + str(time.time())
# Open the HDF5 file (take into account also older TDF versions):
f_in = getHDF5( infile, 'r' )
if "/tomo" in f_in:
dset = f_in['tomo']
else:
dset = f_in['exchange/data']
num_proj = tdf.get_nr_projs(dset)
num_sinos = tdf.get_nr_sinos(dset)
# Get flats and darks from cache or from file:
try:
corrplan = cache2plan(infile, tmppath)
except Exception as e:
#print "Error(s) when reading from cache"
corrplan = extract_flatdark(f_in, True, tmplog)
remove(tmplog)
plan2cache(corrplan, infile, tmppath)
# Get first and the 180 deg projections:
im1 = tdf.read_tomo(dset,proj_from).astype(float32)
idx = int(round( (proj_to - proj_from)/angles * pi)) + proj_from
im2 = tdf.read_tomo(dset,idx).astype(float32)
# Apply simple flat fielding (if applicable):
if (isinstance(corrplan['im_flat_after'], ndarray) and isinstance(corrplan['im_flat'], ndarray) and
isinstance(corrplan['im_dark'], ndarray) and isinstance(corrplan['im_dark_after'], ndarray)) :
im1 = ((abs(im1 - corrplan['im_dark'])) / (abs(corrplan['im_flat'] - corrplan['im_dark'])
+ finfo(float32).eps)).astype(float32)
im2 = ((abs(im2 - corrplan['im_dark_after'])) / (abs(corrplan['im_flat_after'] - corrplan['im_dark_after'])
+ finfo(float32).eps)).astype(float32)
# Scale projections (if required) to get subpixel estimation:
if ( abs(scale - 1.0) > finfo(float32).eps ):
im1 = imresize(im1, (int(round(scale*im1.shape[0])), int(round(scale*im1.shape[1]))), interp='bicubic', mode='F');
im2 = imresize(im2, (int(round(scale*im2.shape[0])), int(round(scale*im2.shape[1]))), interp='bicubic', mode='F');
# Find the center (flipping left-right im2):
cen = findcenter.usecorrelation(im1, im2[ :,::-1])
cen = cen / scale
# Print center to output file:
text_file = open(outfile, "w")
text_file.write(str(int(cen)))
text_file.close()
# Close input HDF5:
f_in.close()
def main(argv): """Extract a 2D image (projection or sinogram) from the input TDF file (DataExchange HDF5) and creates a 32-bit RAW file to disk. Parameters ---------- argv[0] : string The absolute path of the input TDF. argv[1] : int The relative position of the image within the dataset. argv[2] : string One of the following options: 'tomo', 'sino', 'flat', 'dark'. argv[3] : string The absolute path of the output 32-bit RAW image file. Filename will be modified by adding image width, image height, minimum and maximum value of the input TDF dataset. Example ------- tools_extractdata "S:\\dataset.tdf" 128 tomo "R:\\proj" """ try: # # Get input parameters: # infile = argv[0] index = int(argv[1]) imtype = argv[2] outfile = argv[3] # # Body # # Check if file exists: if not os.path.exists(infile): #log = open(logfilename,"a") #log.write(os.linesep + "\tError: input TDF file not found. Process will end.") #log.close() exit() # Open the HDF5 file: f = getHDF5( infile, 'r' ) if (imtype == 'sino'): if "/tomo" in f: dset = f['tomo'] else: dset = f['exchange/data'] im = tdf.read_sino( dset, index ) elif (imtype == 'dark'): if "/dark" in f: dset = f['dark'] else: dset = f['exchange/data_dark'] im = tdf.read_tomo( dset, index ) elif (imtype == 'flat'): if "/flat" in f: dset = f['flat'] else: dset = f['exchange/data_white'] im = tdf.read_tomo( dset, index ) else: if "/tomo" in f: dset = f['tomo'] else: dset = f['exchange/data'] im = tdf.read_tomo( dset, index ) # Remove Infs e NaNs tmp = im[:].astype(numpy.float32) tmp = tmp[numpy.nonzero(numpy.isfinite(tmp))] # Sort the gray levels: tmp = numpy.sort(tmp) # Return as minimum the value the skip 0.30% of "black" tail and 0.05% of "white" tail: low_idx = int(tmp.shape[0] * 0.0030) high_idx = int(tmp.shape[0] * 0.9995) min = tmp[low_idx] max = tmp[high_idx] # Modify file name: outfile = outfile + '_' + str(im.shape[1]) + 'x' + str(im.shape[0]) + '_' + str(min) + '$' + str(max) # Cast type: im = im.astype(float32) # Write RAW data to disk: im.tofile(outfile) except: exit()