nifty.segmentation

Functions

nifty.segmentation.connectedComponents(labels, dense=True, ignoreBackground=False)

get connected components of a label image

Get connected components of an image w.r.t. a 4-neighborhood . This is a high level wrapper for

Parameters:
  • labels (numpy.ndarray) –
  • dense (bool) – should the return labeling be dense (default: {True})
  • ignoreBackground (bool) – should values of zero be excluded (default: {False})
Returns:

[description] [type]
nifty.segmentation.distanceTransformWatersheds(pmap, preBinarizationMedianRadius=1, threshold=0.5, preSeedSigma=0.75)

Superpixels for neuro data as in http://brainiac2.mit.edu/isbi_challenge/

Use raw data and membrane probability maps to generate a over-segmentation suitable for neuro data

Parameters:
  • pmap (numpy.ndarray) – Membrane probability in [0,1].
  • preBinarizationMedianRadius (int) – Radius of diskMedian filter applied to the probability map before binarization. (default:{1})
  • threshold (float) – threshold to binarize probability map before applying the distance transform (default: {0.5})
  • preSeedSigma (float) – smooth the distance transform image before getting the seeds.
Raises:

RuntimeError – if applied to data with wrong dimensionality
nifty.segmentation.localMaxima(image)

get the local maxima of an image

Get the local maxima wrt a 4-neighborhood on an image. For a plateau, all pixels of this plateau are marked as maximum pixel.

Parameters:image (numpy.ndarray) – the input image
Returns:array which is 1 the maximum 0 elsewhere.
Return type:(numpy.ndarray)
nifty.segmentation.localMaximaSeeds(image)

Get seed from local maxima

Get seeds by running connected components on the local maxima. This is a high level wrapper around nifty.segmentation.localMinima() and nifty.segmentation.connectedComponents()

Parameters:image – [description]
Returns:[description] [type]
Raises:RuntimeError – [description]
nifty.segmentation.localMinima(image)

get the local minima of an image

Get the local minima wrt a 4-neighborhood on an image. For a plateau, all pixels of this plateau are marked as minimum pixel.

Parameters:image (numpy.ndarray) – the input image
Returns:array which is 1 the minimum 0 elsewhere.
Return type:(numpy.ndarray)
nifty.segmentation.localMinimaSeeds(image)

Get seed from local minima

Get seeds by running connected components on the local minima. This is a high level wrapper around nifty.segmentation.localMinima() and nifty.segmentation.connectedComponents()

Parameters:image – [description]
Returns:[description] [type]
Raises:RuntimeError – [description]
nifty.segmentation.markBoundaries(image, segmentation, color=None, thin=True)

Mark the boundaries in an image

Mark boundaries in an image.

Warning

The returned image shape is twice as large as the input if this is True.

Parameters:
  • image – the input image
  • segmentation – the segmentation
  • color (tuple) – the edge color(default: {(0,0,0)})
  • thin (bool) – IF true, the image is interpolated and the boundaries are marked in the interpolated image. This will make the output twice as large.
Returns:

image with marked boundaries. Note that

result image has twice as large shape as the input if thin is True.
Return type:

(numpy.ndarray)
nifty.segmentation.randomColormap(size=10000, zeroToZero=False)
nifty.segmentation.seededWatersheds(heightMap, seeds=None, method='node_weighted', acc='max')

Seeded watersheds segmentation

Get a segmentation via seeded watersheds. This is a high level wrapper around nifty.graph.nodeWeightedWatershedsSegmentation() and nifty.graph.nodeWeightedWatershedsSegmentation().

Parameters:
  • heightMap (numpy.ndarray) – height / evaluation map
  • seeds (numpy.ndarray) – Seeds as non zero elements in the array. (default: {nifty.segmentation.localMinimaSeeds(heightMap)})
  • method (str) –

    Algorithm type can be:

    • ”node_weighted”: ordinary node weighted watershed
    • ”edge_weighted”: edge weighted watershed (minimum spanning tree)

    (default: {“max”})

  • acc (str) –

    If method is “edge_weighted”, one needs to specify how to convert the heightMap into an edgeMap. This parameter specificities this method. Allow values are:

    • ’min’ : Take the minimum value of the endpoints of the edge
    • ’max’ : Take the minimum value of the endpoints of the edge
    • ’sum’ : Take the sum of the values of the endpoints of the edge
    • ’prod’ : Take the product of the values of the endpoints of the edge
    • ’interpixel’ : Take the value of the image at the interpixel coordinate in between the two endpoints. To do this the image is resampled to have shape :math: `2 cdot shape -1 `

    (default: {“max”})
Returns:

the segmentation
Return type:

numpy.ndarray
Raises:

RuntimeError – [description]
nifty.segmentation.segmentOverlay(image, segmentation, beta=0.5, zeroToZero=False, showBoundaries=True, color=None, thin=True)
nifty.segmentation.slic(image, nSegments, components)

same as skimage.segmentation.slic