Source code for tomopy.misc.corr

#!/usr/bin/env python
# -*- coding: utf-8 -*-

# #########################################################################
# Copyright (c) 2015-2019, UChicago Argonne, LLC. All rights reserved.    #
#                                                                         #
# Copyright 2015-2019. UChicago Argonne, LLC. This software was produced  #
# under U.S. Government contract DE-AC02-06CH11357 for Argonne National   #
# Laboratory (ANL), which is operated by UChicago Argonne, LLC for the    #
# U.S. Department of Energy. The U.S. Government has rights to use,       #
# reproduce, and distribute this software.  NEITHER THE GOVERNMENT NOR    #
# UChicago Argonne, LLC MAKES ANY WARRANTY, EXPRESS OR IMPLIED, OR        #
# ASSUMES ANY LIABILITY FOR THE USE OF THIS SOFTWARE.  If software is     #
# modified to produce derivative works, such modified software should     #
# be clearly marked, so as not to confuse it with the version available   #
# from ANL.                                                               #
#                                                                         #
# Additionally, redistribution and use in source and binary forms, with   #
# or without modification, are permitted provided that the following      #
# conditions are met:                                                     #
#                                                                         #
#     * Redistributions of source code must retain the above copyright    #
#       notice, this list of conditions and the following disclaimer.     #
#                                                                         #
#     * Redistributions in binary form must reproduce the above copyright #
#       notice, this list of conditions and the following disclaimer in   #
#       the documentation and/or other materials provided with the        #
#       distribution.                                                     #
#                                                                         #
#     * Neither the name of UChicago Argonne, LLC, Argonne National       #
#       Laboratory, ANL, the U.S. Government, nor the names of its        #
#       contributors may be used to endorse or promote products derived   #
#       from this software without specific prior written permission.     #
#                                                                         #
# THIS SOFTWARE IS PROVIDED BY UChicago Argonne, LLC AND CONTRIBUTORS     #
# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT       #
# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS       #
# FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL UChicago     #
# Argonne, LLC OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,        #
# INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,    #
# BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;        #
# LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER        #
# CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT      #
# LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN       #
# ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE         #
# POSSIBILITY OF SUCH DAMAGE.                                             #
# #########################################################################
"""
Module for data correction and masking functions.
"""

from __future__ import absolute_import, division, print_function, unicode_literals

import numpy as np
import scipy.ndimage
import tomopy.util.mproc as mproc
import tomopy.util.dtype as dtype
import tomopy.util.extern as extern
import logging
import warnings
import numexpr as ne
import concurrent.futures as cf
from scipy.signal import medfilt2d

logger = logging.getLogger(__name__)

__author__ = "Doga Gursoy, William Judge"
__credits__ = "Mark Rivers, Xianghui Xiao"
__copyright__ = "Copyright (c) 2015, UChicago Argonne, LLC."
__docformat__ = "restructuredtext en"
__all__ = [
    "adjust_range",
    "circ_mask",
    "gaussian_filter",
    "median_filter",
    "median_filter_cuda",
    "median_filter_nonfinite",
    "median_filter3d",
    "remove_outlier3d",
    "sobel_filter",
    "remove_nan",
    "remove_neg",
    "remove_outlier",
    "remove_outlier1d",
    "remove_outlier_cuda",
    "remove_ring",
    "enhance_projs_aps_1id",
    "inpainter_morph",
]


[docs]def adjust_range(arr, dmin=None, dmax=None): """ Change dynamic range of values in an array. Parameters ---------- arr : ndarray Input array. dmin, dmax : float, optional Mininum and maximum values to rescale data. Returns ------- ndarray Output array. """ if dmax is None: dmax = np.max(arr) if dmin is None: dmin = np.min(arr) if dmax < np.max(arr): arr[arr > dmax] = dmax if dmin > np.min(arr): arr[arr < dmin] = dmin return arr
[docs]def gaussian_filter(arr, sigma=3, order=0, axis=0, ncore=None): """ Apply Gaussian filter to 3D array along specified axis. Parameters ---------- arr : ndarray Input array. sigma : scalar or sequence of scalars Standard deviation for Gaussian kernel. The standard deviations of the Gaussian filter are given for each axis as a sequence, or as a single number, in which case it is equal for all axes. order : {0, 1, 2, 3} or sequence from same set, optional Order of the filter along each axis is given as a sequence of integers, or as a single number. An order of 0 corresponds to convolution with a Gaussian kernel. An order of 1, 2, or 3 corresponds to convolution with the first, second or third derivatives of a Gaussian. Higher order derivatives are not implemented axis : int, optional Axis along which median filtering is performed. ncore : int, optional Number of cores that will be assigned to jobs. Returns ------- ndarray 3D array of same shape as input. """ arr = dtype.as_float32(arr) out = np.empty_like(arr) if ncore is None: ncore = mproc.mp.cpu_count() with cf.ThreadPoolExecutor(ncore) as e: slc = [slice(None)] * arr.ndim for i in range(arr.shape[axis]): slc[axis] = i e.submit( scipy.ndimage.gaussian_filter, arr[tuple(slc)], sigma, order=order, output=out[tuple(slc)], ) return out
[docs]def median_filter(arr, size=3, axis=0, ncore=None): """ Apply median filter to 3D array along specified axis. Parameters ---------- arr : ndarray Input array. size : int, optional The size of the filter. axis : int, optional Axis along which median filtering is performed. ncore : int, optional Number of cores that will be assigned to jobs. Returns ------- ndarray Median filtered 3D array. """ arr = dtype.as_float32(arr) out = np.empty_like(arr) if ncore is None: ncore = mproc.mp.cpu_count() with cf.ThreadPoolExecutor(ncore) as e: slc = [slice(None)] * arr.ndim for i in range(arr.shape[axis]): slc[axis] = i e.submit( scipy.ndimage.median_filter, arr[tuple(slc)], size=(size, size), output=out[tuple(slc)], ) return out
[docs]def median_filter_cuda(arr, size=3, axis=0): """ Apply median filter to 3D array along 0 axis with GPU support. The winAllow is for A6000, Tian X support 3 to 8 Parameters ---------- arr : ndarray Input array. size : int, optional The size of the filter. axis : int, optional Axis along which median filtering is performed. Returns ------- ndarray Median filtered 3D array. Example ------- import tomocuda tomocuda.remove_outlier_cuda(arr, dif, 5) For more information regarding install and using tomocuda, check https://github.com/kyuepublic/tomocuda for more information """ try: import tomocuda winAllow = range(2, 16) if axis != 0: arr = np.swapaxes(arr, 0, axis) if size in winAllow: loffset = int(size / 2) roffset = int((size - 1) / 2) prjsize = arr.shape[0] imsizex = arr.shape[2] imsizey = arr.shape[1] filter = tomocuda.mFilter(imsizex, imsizey, prjsize, size) out = np.zeros(shape=(prjsize, imsizey, imsizex), dtype=np.float32) for step in range(prjsize): # im_noisecu = arr[:][step][:].astype(np.float32) im_noisecu = arr[step].astype(np.float32) im_noisecu = np.lib.pad( im_noisecu, ((loffset, roffset), (loffset, roffset)), "symmetric", ) im_noisecu = im_noisecu.flatten() filter.setCuImage(im_noisecu) filter.run2DFilter(size) results = filter.retreive() results = results.reshape(imsizey, imsizex) out[step] = results if axis != 0: out = np.swapaxes(out, 0, axis) else: warnings.warn("Window size not support, using cpu median filter") out = median_filter(arr, size, axis) except ImportError: warnings.warn("The tomocuda is not support, using cpu median filter") out = median_filter(arr, size, axis) return out
[docs]def median_filter_nonfinite(arr, size=3, callback=None): """ Remove nonfinite values from a 3D array using an in-place 2D median filter. The 2D selective median filter is applied along the last two axes of the array. .. versionadded:: 1.11 Parameters ---------- arr : ndarray The 3D array with nonfinite values in it. size : int, optional The size of the filter. callback : func(total, description, unit) A function called after every internal loop iteration. total is number of loop iterations. description is 'Nonfinite median filter'. unit is ' prjs'. Returns ------- ndarray The corrected 3D array with all nonfinite values removed based upon the local median value defined by the kernel size. Raises ------ ValueError If the filter comes across a kernel only containing non-finite values a ValueError is raised for the user to increase their kernel size. """ # Defining a callback function if None is provided if callback is None: def callback(total, description, unit): pass # Iterating throug each projection to save on RAM for projection in arr: nonfinite_idx = np.nonzero(~np.isfinite(projection)) projection_copy = projection.copy() # Iterating through each bad value and replace it with finite median for x_idx, y_idx in zip(*nonfinite_idx): # Determining the lower and upper bounds for kernel x_lower = max(0, x_idx - (size // 2)) x_higher = min(arr.shape[1], x_idx + (size // 2) + 1) y_lower = max(0, y_idx - (size // 2)) y_higher = min(arr.shape[2], y_idx + (size // 2) + 1) # Extracting kernel data and fining finite median kernel_cropped_arr = projection_copy[x_lower:x_higher, y_lower:y_higher] if len(kernel_cropped_arr[np.isfinite(kernel_cropped_arr)]) == 0: raise ValueError( "Found kernel containing only non-finite values.\ Please increase kernel size" ) median_corrected_arr = np.median( kernel_cropped_arr[np.isfinite(kernel_cropped_arr)] ) # Replacing bad data with finite median projection[x_idx, y_idx] = median_corrected_arr callback(arr.shape[0], "Nonfinite median filter", " prjs") return arr
[docs]def median_filter3d(arr, size=3, ncore=None): """ Apply 3D median filter to a 3D array. .. versionadded:: 1.13 Parameters ---------- arr : ndarray Input 3D array either float32 or uint16 data type. size : int, optional The size of the filter's kernel. ncore : int, optional Number of cores that will be assigned to jobs. All cores will be used if unspecified. Returns ------- ndarray Median filtered 3D array either float32 or uint16 data type. Raises ------ ValueError If the input array is not three dimensional. """ if ncore is None: ncore = mproc.mp.cpu_count() input_type = arr.dtype if (input_type != "float32") and (input_type != "uint16"): arr = dtype.as_float32(arr) # silent convertion to float32 data type out = np.empty_like(arr) dif = 0.0 # set to 0 to avoid selective filtering # convert the full kernel size (odd int) to a half size as the C function requires it kernel_half_size = (max(int(size), 3) - 1) // 2 if arr.ndim == 3: dz, dy, dx = arr.shape if (dz == 0) or (dy == 0) or (dx == 0): raise ValueError("The length of one of dimensions is equal to zero") else: raise ValueError("The input array must be a 3D array") # perform full 3D filtering if input_type == "float32": extern.c_median_filt3d_float32( np.ascontiguousarray(arr), out, kernel_half_size, dif, ncore, dx, dy, dz ) else: extern.c_median_filt3d_uint16( np.ascontiguousarray(arr), out, kernel_half_size, dif, ncore, dx, dy, dz ) return out
[docs]def remove_outlier3d(arr, dif, size=3, ncore=None): """ Selectively applies 3D median filter to a 3D array to remove outliers. Also called a dezinger. .. versionadded:: 1.13 Parameters ---------- arr : ndarray Input 3D array either float32 or uint16 data type. dif : float Expected difference value between outlier value and the median value of the array. size : int, optional The size of the filter's kernel. ncore : int, optional Number of cores that will be assigned to jobs. All cores will be used if unspecified. Returns ------- ndarray Dezingered 3D array either float32 or uint16 data type. Raises ------ ValueError If the input array is not three dimensional. """ if ncore is None: ncore = mproc.mp.cpu_count() input_type = arr.dtype if (input_type != "float32") and (input_type != "uint16"): arr = dtype.as_float32(arr) # silent convertion to float32 data type out = np.copy(arr, order="C") # convert the full kernel size (odd int) to a half size as the C function requires it kernel_half_size = (max(int(size), 3) - 1) // 2 if arr.ndim == 3: dz, dy, dx = arr.shape if (dz == 0) or (dy == 0) or (dx == 0): raise ValueError("The length of one of dimensions is equal to zero") else: raise ValueError("The input array must be a 3D array") # perform full 3D filtering if input_type == "float32": extern.c_median_filt3d_float32( np.ascontiguousarray(arr), out, kernel_half_size, dif, ncore, dx, dy, dz ) else: extern.c_median_filt3d_uint16( np.ascontiguousarray(arr), out, kernel_half_size, dif, ncore, dx, dy, dz ) return out
[docs]def sobel_filter(arr, axis=0, ncore=None): """ Apply Sobel filter to 3D array along specified axis. Parameters ---------- arr : ndarray Input array. axis : int, optional Axis along which sobel filtering is performed. ncore : int, optional Number of cores that will be assigned to jobs. Returns ------- ndarray 3D array of same shape as input. """ arr = dtype.as_float32(arr) out = np.empty_like(arr) if ncore is None: ncore = mproc.mp.cpu_count() with cf.ThreadPoolExecutor(ncore) as e: slc = [slice(None)] * arr.ndim for i in range(arr.shape[axis]): slc[axis] = i e.submit(filters.sobel, arr[slc], output=out[slc]) return out
[docs]def remove_nan(arr, val=0.0, ncore=None): """ Replace NaN values in array with a given value. Parameters ---------- arr : ndarray Input array. val : float, optional Values to be replaced with NaN values in array. ncore : int, optional Number of cores that will be assigned to jobs. Returns ------- ndarray Corrected array. """ arr = dtype.as_float32(arr) val = np.float32(val) with mproc.set_numexpr_threads(ncore): ne.evaluate("where(arr!=arr, val, arr)", out=arr) return arr
[docs]def remove_neg(arr, val=0.0, ncore=None): """ Replace negative values in array with a given value. Parameters ---------- arr : ndarray Input array. val : float, optional Values to be replaced with negative values in array. ncore : int, optional Number of cores that will be assigned to jobs. Returns ------- ndarray Corrected array. """ arr = dtype.as_float32(arr) val = np.float32(val) with mproc.set_numexpr_threads(ncore): ne.evaluate("where(arr<0, val, arr)", out=arr) return arr
[docs]def remove_outlier(arr, dif, size=3, axis=0, ncore=None, out=None): """ Remove high intensity bright spots from a N-dimensional array by chunking along the specified dimension, and performing (N-1)-dimensional median filtering along the other dimensions. Parameters ---------- arr : ndarray Input array. dif : float Expected difference value between outlier value and the median value of the array. size : int Size of the median filter. axis : int, optional Axis along which to chunk. ncore : int, optional Number of cores that will be assigned to jobs. out : ndarray, optional Output array for result. If same as arr, process will be done in-place. Returns ------- ndarray Corrected array. """ tmp = np.empty_like(arr) ncore, chnk_slices = mproc.get_ncore_slices(arr.shape[axis], ncore=ncore) filt_size = [size] * arr.ndim filt_size[axis] = 1 with cf.ThreadPoolExecutor(ncore) as e: slc = [slice(None)] * arr.ndim for i in range(ncore): slc[axis] = chnk_slices[i] e.submit( scipy.ndimage.median_filter, arr[tuple(slc)], size=filt_size, output=tmp[tuple(slc)], ) arr = dtype.as_float32(arr) tmp = dtype.as_float32(tmp) dif = np.float32(dif) with mproc.set_numexpr_threads(ncore): out = ne.evaluate("where(arr-tmp>=dif,tmp,arr)", out=out) return out
[docs]def remove_outlier1d(arr, dif, size=3, axis=0, ncore=None, out=None): """ Remove high intensity bright spots from an array, using a one-dimensional median filter along the specified axis. Parameters ---------- arr : ndarray Input array. dif : float Expected difference value between outlier value and the median value of the array. size : int Size of the median filter. axis : int, optional Axis along which median filtering is performed. ncore : int, optional Number of cores that will be assigned to jobs. out : ndarray, optional Output array for result. If same as arr, process will be done in-place. Returns ------- ndarray Corrected array. """ arr = dtype.as_float32(arr) dif = np.float32(dif) tmp = np.empty_like(arr) other_axes = [i for i in range(arr.ndim) if i != axis] largest = np.argmax([arr.shape[i] for i in other_axes]) lar_axis = other_axes[largest] ncore, chnk_slices = mproc.get_ncore_slices(arr.shape[lar_axis], ncore=ncore) filt_size = [1] * arr.ndim filt_size[axis] = size with cf.ThreadPoolExecutor(ncore) as e: slc = [slice(None)] * arr.ndim for i in range(ncore): slc[lar_axis] = chnk_slices[i] e.submit( scipy.ndimage.median_filter, arr[slc], size=filt_size, output=tmp[slc], mode="mirror", ) with mproc.set_numexpr_threads(ncore): out = ne.evaluate("where(arr-tmp>=dif,tmp,arr)", out=out) return out
[docs]def remove_outlier_cuda(arr, dif, size=3, axis=0): """ Remove high intensity bright spots from a 3D array along axis 0 dimension using GPU. Parameters ---------- arr : ndarray Input array. dif : float Expected difference value between outlier value and the median value of the array. size : int Size of the median filter. axis : int, optional Axis along which outlier removal is performed. Returns ------- ndarray Corrected array. Example ------- >>> import tomocuda >>> tomocuda.remove_outlier_cuda(arr, dif, 5) For more information regarding install and using tomocuda, check https://github.com/kyuepublic/tomocuda for more information """ arr = dtype.as_float32(arr) dif = np.float32(dif) try: import tomocuda winAllow = range(2, 16) if axis != 0: arr = np.swapaxes(arr, 0, axis) if size in winAllow: prjsize = arr.shape[0] loffset = int(size / 2) roffset = int((size - 1) / 2) imsizex = arr.shape[2] imsizey = arr.shape[1] filter = tomocuda.mFilter(imsizex, imsizey, prjsize, size) out = np.zeros(shape=(prjsize, imsizey, imsizex), dtype=np.float32) for step in range(prjsize): im_noisecu = arr[step].astype(np.float32) im_noisecu = np.lib.pad( im_noisecu, ((loffset, roffset), (loffset, roffset)), "symmetric" ) im_noisecu = im_noisecu.flatten() filter.setCuImage(im_noisecu) filter.run2DRemoveOutliner(size, dif) results = filter.retreive() results = results.reshape(imsizey, imsizex) out[step] = results if axis != 0: out = np.swapaxes(out, 0, axis) else: warnings.warn("Window size not support, using cpu outlier removal") out = remove_outlier(arr, dif, size) except ImportError: warnings.warn("The tomocuda is not support, using cpu outlier removal") out = remove_outlier(arr, dif, size) return out
[docs]def remove_ring( rec, center_x=None, center_y=None, thresh=300.0, thresh_max=300.0, thresh_min=-100.0, theta_min=30, rwidth=30, int_mode="WRAP", ncore=None, nchunk=None, out=None, ): """ Remove ring artifacts from images in the reconstructed domain. Descriptions of parameters need to be more clear for sure. Parameters ---------- arr : ndarray Array of reconstruction data center_x : float, optional abscissa location of center of rotation center_y : float, optional ordinate location of center of rotation thresh : float, optional maximum value of an offset due to a ring artifact thresh_max : float, optional max value for portion of image to filter thresh_min : float, optional min value for portion of image to filer theta_min : int, optional Features larger than twice this angle (degrees) will be considered a ring artifact. Must be less than 180 degrees. rwidth : int, optional Maximum width of the rings to be filtered in pixels int_mode : str, optional 'WRAP' for wrapping at 0 and 360 degrees, 'REFLECT' for reflective boundaries at 0 and 180 degrees. ncore : int, optional Number of cores that will be assigned to jobs. nchunk : int, optional Chunk size for each core. out : ndarray, optional Output array for result. If same as arr, process will be done in-place. Returns ------- ndarray Corrected reconstruction data """ rec = dtype.as_float32(rec) if out is None: out = rec.copy() else: out = dtype.as_float32(out) dz, dy, dx = rec.shape if center_x is None: center_x = (dx - 1.0) / 2.0 if center_y is None: center_y = (dy - 1.0) / 2.0 if int_mode.lower() == "wrap": int_mode = 0 elif int_mode.lower() == "reflect": int_mode = 1 else: raise ValueError("int_mode should be WRAP or REFLECT") if not 0 <= theta_min < 180: raise ValueError("theta_min should be in the range [0 - 180)") args = ( center_x, center_y, dx, dy, dz, thresh_max, thresh_min, thresh, theta_min, rwidth, int_mode, ) axis_size = rec.shape[0] ncore, nchunk = mproc.get_ncore_nchunk(axis_size, ncore, nchunk) with cf.ThreadPoolExecutor(ncore) as e: for offset in range(0, axis_size, nchunk): slc = np.s_[offset : offset + nchunk] e.submit(extern.c_remove_ring, out[slc], *args) return out
[docs]def circ_mask(arr, axis, ratio=1, val=0.0, ncore=None): """ Apply circular mask to a 3D array. Parameters ---------- arr : ndarray Arbitrary 3D array. axis : int Axis along which mask will be performed. ratio : int, optional Ratio of the mask's diameter in pixels to the smallest edge size along given axis. val : int, optional Value for the masked region. Returns ------- ndarray Masked array. """ arr = dtype.as_float32(arr) val = np.float32(val) _arr = arr.swapaxes(0, axis) dx, dy, dz = _arr.shape mask = _get_mask(dy, dz, ratio) with mproc.set_numexpr_threads(ncore): ne.evaluate("where(mask, _arr, val)", out=_arr) return _arr.swapaxes(0, axis)
def _get_mask(dx, dy, ratio): """ Calculate 2D boolean circular mask. Parameters ---------- dx, dy : int Dimensions of the 2D mask. ratio : int Ratio of the circle's diameter in pixels to the smallest mask dimension. Returns ------- ndarray 2D boolean array. """ rad1 = dx / 2.0 rad2 = dy / 2.0 if dx < dy: r2 = rad1 * rad1 else: r2 = rad2 * rad2 y, x = np.ogrid[0.5 - rad1 : 0.5 + rad1, 0.5 - rad2 : 0.5 + rad2] return x * x + y * y < ratio * ratio * r2
[docs]def enhance_projs_aps_1id(imgstack, median_ks=5, ncore=None): """ Enhance the projection images with weak contrast collected at APS 1ID This filter uses a median fileter (will be switched to enhanced recursive median fileter, ERMF, in the future) for denoising, and a histogram equalization for dynamic range adjustment to bring out the details. Parameters ---------- imgstack : np.ndarray tomopy images stacks (axis_0 is the oemga direction) median_ks : int, optional 2D median filter kernel size for local noise suppresion ncore : int, optional number of cores used for speed up Returns ------- ndarray 3D enhanced image stacks. """ ncore = mproc.mp.cpu_count() - 1 if ncore is None else ncore # need to use multiprocessing to speed up the process tmp = [] with cf.ProcessPoolExecutor(ncore) as e: for n_img in range(imgstack.shape[0]): tmp.append( e.submit( _enhance_img, imgstack[n_img, :, :], median_ks, ) ) return np.stack([me.result() for me in tmp], axis=0)
def _enhance_img(img, median_ks, normalized=True): """ Enhance the projection image from aps 1ID to counter its weak contrast nature Parameters ---------- img : ndarray original projection image collected at APS 1ID median_ks: int kernel size of the 2D median filter, must be odd normalized: bool, optional specify whether the enhanced image is normalized between 0 and 1, default is True Returns ------- ndarray enhanced projection image """ wgt = _calc_histequal_wgt(img) img = medfilt2d(img, kernel_size=median_ks).astype(np.float64) img = ne.evaluate("(img**2)*wgt", out=img) return img / img.max() if normalized else img def _calc_histequal_wgt(img): """ Calculate the histogram equalization weight for a given image Parameters ---------- img : ndarray 2D images Returns ------- ndarray histogram euqalization weights (0-1) in the same shape as original image """ return (np.sort(img.flatten()).searchsorted(img) + 1) / np.prod(img.shape)
[docs]def inpainter_morph( arr, mask, size=5, iterations=2, inpainting_type="random", axis=None, ncore=None, ): """ Apply 3D or 2D morphological inpainter (extrapolator) to a given missing data array using provided mask (boolean). The algorithm is presented in the paper :cite:`Kazantsev:23`. If applied to 3D tomographic data, one can use the 2D module applied to a sinogram by specifing the axis in the list of parameters bellow. Alternatively, one can use the 3D module with symmetric 3D kernels (axis=None). The latter is more recommended for 3D data as it ensures the smoother intensity transition in every direction. .. versionadded:: 1.15 Parameters ---------- arr : ndarray Input 3D or 2D array of float32 data type with the missing data. mask : ndarray, bool Boolean mask array of the same size as arr. True values in the mask indicate the region that needs to be inpainted. size : int, optional The size of the searching window (a kernel). iterations : int, optional Iterations of the algorithm after the inpainted region was processed fully. The larger numbers usually lead to oversmoothing of the area, we recommend 2. inpainting_type : str, optional The type of the inpainting technique. Choose between 'mean', 'median' or 'random' neighbour selection. We suggest using 'random' to minimise the "spilling" of intensities inside the inpainted areas. axis : int, optional Split the data along this axis, and perform inpainting in the remaining dimensions. If set to `None` then the 3D inpainting is enabled, which is recommended to use for 3D data. If the 2D inpainting used for 3D data, it is best to apply it in sinogram space. ncore : int, optional Number of cores that will be assigned to jobs. All cores will be used if unspecified. Returns ------- ndarray Inpainted array of float32 data type. Raises ------ ValueError If the input array is not float32. If the input mask array is not bool. If inpainting_type is not 'mean', 'median' or 'random'. If axis is an integer for 2D input data. If axis is outside of the accepted range. """ if ncore is None: ncore = mproc.mp.cpu_count() if arr.dtype != "float32": arr = dtype.as_float32(arr) # silent convertion to float32 data type if mask.dtype != "bool": raise ValueError( "Please make sure that the mask provided is boolean where true values define the missing data regions!" ) out = np.empty_like(arr, order="C") if inpainting_type not in ["mean", "median", "random"]: raise ValueError( "Inpainting type should be chosen as 'mean', 'median' or 'random'" ) if inpainting_type == "mean": inp_type_int = 0 if inpainting_type == "median": inp_type_int = 1 if inpainting_type == "random": inp_type_int = 2 if axis is not None: if arr.ndim == 2: raise ValueError("The axis number is valid only for 3D data, please set axis to None for 2D data") else: if axis < 0 or axis > 2: raise ValueError("The accepted axes range for 3D data is [0:2]") if arr.ndim == 3: dz, dy, dx = arr.shape if (dz == 0) or (dy == 0) or (dx == 0): raise ValueError("The length of one of dimensions is equal to zero") if axis is not None: # perform 2d inpainting for 3d data using the provided axis slices_list = [slice(None), slice(None), slice(None)] for m in range(arr.shape[axis]): slices_list[axis] = slice(m,m+1) slice2d = arr[tuple(slices_list)].squeeze() mask2d = mask[tuple(slices_list)].squeeze() dy, dx = slice2d.shape out2d = np.empty_like(slice2d, order="C") extern.c_inpainter( np.ascontiguousarray(slice2d), np.ascontiguousarray(mask2d), out2d, iterations, size, inp_type_int, ncore, dx, dy, 1, ) if axis == 0: out[m, :, :] = out2d elif axis == 1: out[:, m, :] = out2d else: out[:, :, m] = out2d else: # 3d inpaiting for 3d data extern.c_inpainter( np.ascontiguousarray(arr), np.ascontiguousarray(mask), out, iterations, size, inp_type_int, ncore, dx, dy, dz, ) else: dy, dx = arr.shape if (dy == 0) or (dx == 0): raise ValueError("The length of one of dimensions is equal to zero") extern.c_inpainter( np.ascontiguousarray(arr), np.ascontiguousarray(mask), out, iterations, size, inp_type_int, ncore, dx, dy, 1, ) return out