Regional extrema

Regional extrema detection module.

Detection of local minima in a membrane marked intensity image, allows finding seeds for the watershed algorithm. Detection of local maxima in a nuclei marked intensity image, allows finding nuclei position in the image.

timagetk.algorithms.regionalext.regional_extrema(image, height, method='minima', **kwargs)[source]

Regional extrema detection algorithm.

Valid method values are:

  • minima, for detection of connected regional minima

  • maxima, for detection of connected regional maxima

Parameters:

  • image (timagetk.SpatialImage or timagetk.MultiChannelImage) – Input image to use for regional extrema detection.

  • height (int or float) – The height of the regional extrema to detect.

  • method ({'minima', 'maxima'}) – The method to use for regional extrema detection, DEF_RE_METHOD by default.

  • channel (str) – If a MultiChannelImage is used as input image, select the channel to use with this algorithm.

  • connectivity ({4, 6, 8, 10, 18, 26}) – The connectivity of the structuring elements, DEF_CONNECTIVITY by default. Connectivity is among the 4-, 6-, 8-, 18-, 26-neighborhoods. 4 and 8 are for 2D images, the others for 3D images.

  • params (str) – CLI parameter string used by vt.cell_filter method.

Returns:

LabelledImage – Image with labelled regional minima or maxima.

Raises:

  • TypeError – If image is not a LabelledImage.

  • ValueError – If the image returned by vt.regionalext is None.

See also

timagetk.third_party.vt_parser.REGIONALEXT_METHODS, timagetk.third_party.vt_parser.DEF_RE_METHOD, timagetk.third_party.vt_parser.DEF_CONNECTIVITY, timagetk.third_party.vt_parser.regionalext_kwargs, timagetk.third_party.vt_parser.general_kwargs, timagetk.third_party.vt_parser.parallel_kwargs

Notes

A regional minimum, resp. maximum, M of an image I at elevation t is a connected component of pixels (or voxels) with the value t whose external boundary pixels (or voxels) have a value strictly greater, resp. less, than t [Soille]. The result image is then a binary image, where 1 indicate a regional minimum, resp. maximum.

Providing a contrast criterion h remove the regional minima, resp. maxima, whose depth is higher, resp. lower, to this threshold level h. It is then called a h-minima transformation, resp. h-maxima. The result image is then a height image, i.e. the pixels (or voxels) values correspond to the depth of the regional extrema.

References

Examples

>>> import numpy as np
>>> from timagetk.io.util import shared_dataset
>>> from timagetk.io import imread
>>> from timagetk.algorithms.regionalext import regional_extrema
>>> from timagetk.visu.mplt import grayscale_imshow

>>> # Example #1 - height-5 local minima detection on 3D image:
>>> image = imread(shared_dataset("p58")[0])
>>> regmin_img = regional_extrema(image, height=3, method='minima')
>>> np.unique(regmin_img)

>>> # Example #2 - local minima detection on 2D image for various height values:
>>> zslice = 10
>>> subimg = image.get_slice(zslice, 'z')
>>> hmins = [10, 25, 50]
>>> n_hmins = len(hmins)
>>> extmin_imgs = {hmin: regional_extrema(subimg, hmin, "minima", connectivity=8) for hmin in hmins}
>>> images = [subimg] + [extmin_imgs[hmin] for hmin in hmins]
>>> titles = ["Original (z-slice: {})".format(zslice)] + ["H-minima (h: {})".format(hmin) for hmin in hmins]
>>> grayscale_imshow(images, title=titles, val_range=['type']+['auto']*n_hmins, cmap=['gray']+['viridis']*n_hmins)