porespy.networks.magnet#

porespy.networks.magnet(im, sk=None, parallel_kw=None, surface=False, voxel_size=1, s=None, l_max=7, throat_junctions=None, throat_area=False, **kwargs)#

Perform a Medial Axis Guided Network ExtracTion (MAGNET) on an image of porous media.

This is a modernized python implementation of a classical network extraction method. First, the skeleton of the provided image is determined. The skeleton can be computed in serial or parallel modes. Next, all the junction points of the skeleton are determined by using convolution including terminal points on the ends of branches. ClustersPores are then inserted at these points. The size of the pores inserted is based on the distance transform value at it’s junction. This approach results in many long throats so more pores are added using a maximum filter along long throats to find openings. To ensure an efficient network extraction method, only the most fundamential pore and throat properties are returned.

Parameters:

  • im (ndarray) – An image of the porous material of interest. Be careful of floating solids in the 3D image as this will result in a hollow shell after taking the skeleton. Floating solids are removed from the image by default prior to taking the skeleton.

  • sk (ndarray) – Optionally provide your own skeleton of the image. If sk is None the skeleton is computed using skimage.morphology.skeleton_3d. A check is made to ensure no shells are found in the resulting skeleton.

  • surface (boolean) – If False disconnected solid at the surface of the image is NOT trimmed. This is the default mode. However, if True, disconnected solid at the surface of the image is trimmed. This is NOT applied when im is 2d.

  • parallel_kw (dict) – Dictionary containing the settings for parallelization by chunking. If None is provided, parallelization does not occur. The default is None. The optional settings include divs (scalar or list of scalars, default = [2, 2, 2]), overlap (scalar or list of scalars, optional), and cores (scalar, default is all available cores). See documentaion on ps.networks.skeleton for more information.

  • voxel_size (scalar (default = 1)) – The resolution of the image, expressed as the length of one side of a voxel, so the volume of a voxel would be voxel_size-cubed

  • s (int (default = None)) – The hard threshold for determining “near” junctions. If None is passed, then the distance transform is used to merge junctions. The default is None.

  • l_max (scalar (default = 7)) – The size of the maximum filter used in finding junction along long throats. This argument is only used when throat_junctions is set to “maximum filter” mode.

  • throat_junctions (str) – The mode to use when finding throat junctions. The options are “maximum filter” or “fast marching”. If None is given, then throat junctions are not found (this is the default).

  • throat_area (boolean (default = FALSE)) – Set this argument to TRUE to calculate throat area using get_throat_area. The area is calculated at the throat voxel with the minimum distance transform value. If TRUE, an equivalent throat diameter is returned. The default value is FALSE, for computational efficiency sake. See get_throat_area() documentation for more information.

Returns:

results – A custom object with the following data added as named attributes: ‘network’ A dictionary containing the most important pore and throat size data and topological data. ‘sk’ The skeleton of the image is also returned. ‘juncs’ An ndarray the same shape as im with clusters of junction voxels not uniquely labelled. ‘throat_area’ If throat_area argument is set to FALSE (default), then None is returned. However, if throat_area is set to TRUE, then the measured throat area from get_throat_area is returned here.

Return type:

Results object

On this page