perch.ph

Classes

`PH`()	Base persistent homology class.
`Structures`([structures, img_shape, wcs, ...])	Class for managing a collection of Structure objects.

class perch.ph.PH

Bases: object

Base persistent homology class.

Attributes:

datanp.ndarray: Input data array.
n_dimint: Number of dimensions in image.
max_Hiint: Maximum homology dimension to compute.
generatorsnp.ndarray: Persistent homology generators.
strucsperch.Structures: Structure object.
ph_fxnfunction: Function for computing PH (cripser or pycripser).
noisefloat: Noise map.

compute_hom(max_Hi=None, wcs=None, flip_data=True, verbose=True, embedded=False, noise=None, prep_img_kwargs={})

Compute persistent homology.

Parameters:

datanp.ndarray: Input data array.
max_Hiint: Maximum homology dimension to compute.
wcsastropy.wcs.WCS: WCS object.
flip_databool: Flip data array.
verbosebool: Print progress.
embeddedbool: Compute embedded PH.
noisenp.ndarray: Noise map of same shape as data.

Returns:

perch.PH: Persistent homology object.

Examples

Compute persistent homology of a small 2D image with a single Gaussian peak:

>>> import numpy as np
>>> from perch.ph import PH
>>> y, x = np.indices((8, 8))
>>> img = np.exp(-((y - 3)**2 + (x - 4)**2) / 2.0).astype(np.float32)
>>> ph = PH.compute_hom(data=img, verbose=False)
>>> ph.generators.shape
(1, 10)

export_generators(fname, odir='./')

Export generators to file.

Parameters:

fnamestr: File name.
odirstr: Output directory.

filter(dimension=None, min_life=None, max_life=None, min_birth=None, max_birth=None, min_death=None, max_death=None, min_life_norm_birth=None, min_life_norm_death=None, inds_dir=None, mask=None)

Filter structures.

Parameters:

dimensionint: Homology dimension.
min_lifefloat: Minimum lifetime.
max_lifefloat: Maximum lifetime.
min_birthfloat: Minimum birth.
max_birthfloat: Maximum birth.
min_deathfloat: Minimum death.
max_deathfloat: Maximum death.
min_life_norm_birthfloat: Minimum normalized lifetime at birth.
min_life_norm_deathfloat: Minimum normalized lifetime at death.
inds_dirstr: Directory for saving/loading indices.
masknp.ndarray: Mask for filtering.

Returns:

perch.Structures: Filtered structures.

Examples

Keep only the H0 generators from a precomputed persistent-homology result:

>>> import numpy as np
>>> from perch.ph import PH
>>> y, x = np.indices((8, 8))
>>> img = np.exp(-((y - 3)**2 + (x - 4)**2) / 2.0).astype(np.float32)
>>> ph = PH.compute_hom(data=img, verbose=False)
>>> h0 = ph.filter(dimension=0)
>>> h0.n_struc
1

load_from(odir='./', data=None, wcs=None, max_Hi=None, conv_fac=None, noise=None)

Load generators from file.

Parameters:

fnamestr: File name.
odirstr: Output directory.
datanp.ndarray: Input data array.
wcsastropy.wcs.WCS: WCS object.
max_Hiint: Maximum homology dimension to compute.
conv_facfloat: Conversion factor.
noisenp.ndarray: Noise map of same shape as data.

Returns:

perch.PH: Persistent homology object.