Profiles¶

This submodule contains the attribute profiles related classes.

Example

>>> import sap
>>> import numpy as np

>>> image = np.arange(5*5).reshape(5, 5)

Create the attribute profiles (AP) of image based on area attribute and three thresholds.

>>> aps = sap.attribute_profiles(image, {'area': [10, 100, 1000]})
>>> aps.vectorize()
[[...]]

Create the extended AP of image based on area compactness and volume attributes.

>>> attributes = {'compactness': [.1, .5, .7], 'volume': [10, 100]}
>>> eaps = sap.attribute_profiles(image, attributes)
>>> eaps.vectorize()
[[...]]

Concatenation of profiles to create complex extended profiles.

>>> profiles = sap.attribute_profiles(image, {'area': [10, 100]}) \
...            + sap.feature_profiles(image, {'compactness': [.3, .7]}) \
...            + sap.self_dual_attribute_profiles(image, {'height': [5, 15]})
Profiles[{'attribute': 'area',
  'filtering rule': 'direct',
  'image': -7518820387991786804,
  'name': 'attribute profiles',
  'out feature': 'altitude',
  'profiles': [{'operation': 'thinning', 'threshold': 100},
               {'operation': 'thinning', 'threshold': 10},
               {'operation': 'copy feature altitude'},
               {'operation': 'thickening', 'threshold': 10},
               {'operation': 'thickening', 'threshold': 100}]},
 {'attribute': 'compactness',
  'filtering rule': 'direct',
  'image': -7518820387991786804,
  'name': 'feature profiles',
  'out feature': 'compactness',
  'profiles': [{'operation': 'thinning', 'threshold': 0.7},
               {'operation': 'thinning', 'threshold': 0.3},
               {'operation': 'copy feature compactness'},
               {'operation': 'thickening', 'threshold': 0.3},
               {'operation': 'thickening', 'threshold': 0.7}]},
 {'attribute': 'height',
  'filtering rule': 'direct',
  'image': -7518820387991786804,
  'name': 'self dual attribute profiles',
  'out feature': 'altitude',
  'profiles': [{'operation': 'copy feature altitude'},
               {'operation': 'sd filtering', 'threshold': 5},
               {'operation': 'sd filtering', 'threshold': 15}]}]

class sap.profiles.Profiles(data, description)[source]¶

Bases: object

Base class for profiles.

Parameters

data (list of ndarray) – List of ndarray representing profiles grouped by image or attribute filtering.
description (list of dict) – List of dictionary containing the metadata of the profiles.

diff()[source]¶

Compute the differential of profiles.

Refer to differential() for full documentation.

Returns: differential – The processed differential profiles.
Return type: Profiles

lf(self, local_feature=(np.mean, np.std), patch_size=7)[source]¶

Compute the local features of profiles

Refer to local_features() for full documentation.

local_featurefunction or tuple of functions: The function(s) to describe the local patches.
patch_sizeint: The size of the patches.

Returns: local_features – The local features of profiles.
Return type: Profiles

strip(lambda x: x['operation'] != 'thinning')[source]¶

Remove profiles according to condition. Iteration is done on profiles description.

Refer to strip_profiles() for full documentation.

Parameters: condition (function) – The function (or lambda function) to use on profiles description to filter the profiles.
Returns: new_profiles – Filtered profiles.
Return type: Profiles

See also

strip_profiles(): equivalent function

strip_copy()[source]¶

Remove all the copied images in profiles.

Refer to strip_profiles_copy() for full documentation.

Parameters: profiles (Profiles) – The profiles to strip on the copied images.
Returns: new_profiles – Copy of profiles without copied image.
Return type: Profiles

See also

strip_profiles_copy(): equivalent function

vectorize()[source]¶

Return the vectors of the profiles.

Refer to vectorize() for full documentation.

Returns: vectors – The vectors of the profiles.
Return type: numpy.ndarray

See also

vectorize(): equivalent function.

sap.profiles.alpha_profiles(image, attribute, adjacency=4, image_name=None, filtering_rule='direct')[source]¶

Compute the alpha profiles of an image.

Parameters

image (ndarray) – The image
attribute (dict) – Dictionary of attribute (as key, str) with according thresholds (as values, number).
adjacency (int) – Adjacency used for the tree construction. Default is 4.
image_name (str) – The name of the image (optional). Useful to track filtering process and display. If not set, the name is replaced by the hash of the image.
filtering_rule (str, optional) – The filtering rule to use. It can be ‘direct’, ‘min’, ‘max’ or ‘subtractive’. Default is ‘direct’.

Examples

>>> image = np.arange(5 * 5).reshape(5, 5)
>>> sap.alpha_profiles(image, {'area': [10, 100]})
Profiles{'attribute': 'area',
 'filtering rule': 'direct',
 'name': 'alpha profiles',
 'out feature': 'altitude',
 'profiles': [{'operation': 'copy feature altitude'},
              {'operation': 'alpha filtering', 'threshold': 10},
              {'operation': 'alpha filtering', 'threshold': 100}],
 'tree': {'adjacency': 4, 'image_hash': '44f17c0f', 'image_name': None}}

See also

sap.trees.available_attributes(): List available attributes.

sap.profiles.attribute_profiles(image, attribute, adjacency=4, image_name=None, filtering_rule='direct')[source]¶

Compute the attribute profiles of an image.

Parameters

image (ndarray) – The image
attribute (dict) – Dictionary of attribute (as key, str) with according thresholds (as values, number).
adjacency (int) – Adjacency used for the tree construction. Default is 4.
image_name (str) – The name of the image (optional). Useful to track filtering process and display. If not set, the name is replaced by the hash of the image.
filtering_rule (str, optional) – The filtering rule to use. It can be ‘direct’, ‘min’, ‘max’ or ‘subtractive’. Default is ‘direct’.

Examples

>>> image = np.arange(5*5).reshape(5,5)

>>> sap.attribute_profiles(image, {'area': [10, 100]})
Profiles{'attribute': 'area',
 'filtering rule': 'direct',
 'image': -7518820387991786804,
 'name': 'attribute profiles',
 'out feature': 'altitude',
 'profiles': [{'operation': 'thinning', 'threshold': 100},
              {'operation': 'thinning', 'threshold': 10},
              {'operation': 'copy feature altitude'},
              {'operation': 'thickening', 'threshold': 10},
              {'operation': 'thickening', 'threshold': 100}]}

See also

sap.trees.available_attributes(): List available attributes.

sap.profiles.concatenate((profiles_1, profiles_2, ...))[source]¶

Concatenate a sequence of profiles.

Parameters: sequence (sequence of Profiles) – The sequence of profiles to concatenate.
Returns: profiles – The concatenated profiles.
Return type: Profiles

Examples

>>> aps_a = sap.attribute_profiles(image, {'area': [10, 100]})
>>> aps_b = sap.attribute_profiles(image, {'compactness': [.1, .5]})

>>> aps = sap.concatenate((aps_a, aps_b))

>>> len(aps) == len(aps_a) + len(aps_b)
True

sap.profiles.create_profiles(tree, attribute, out_feature='altitude', filtering_rule='direct', profiles_name='unknow')[source]¶

Compute the profiles of an images. Generic function.

Parameters

image (ndarray) – The image to be profiled.
attribute (dict) – Dictionary of attribute (as key, str) with according thresholds (as values, iterable of thresholds).
tree_type (sap.trees.Tree, serie of sap.trees.Tree) – Tree or pair of tree for non dual filtering (e.g. min-tree and max-tree for attribute profiles).
adjacency (int, optional) – Adjacency used for the tree construction. Default is 4.
image_name (str, optional) – The name of the image Useful to track filtering process and display. If not set, the name is replaced by the hash of the image.
out_feature (str or list, optional) – Out feature of the profiles. Can be ‘altitude’ (default), ‘same’ or a list of feature. If ‘same’ then out feature of the profiles match the filtering attribute. Refer to feature_profiles() and self_dual_feature_profiles() for more details.
filtering_rule (str, optional) – The filtering rule to use. It can be ‘direct’, ‘min’, ‘max’ or ‘subtractive’. Default is ‘direct’.
profiles_name (str, optional) – Name of the profiles (e.g. ‘attribute profiles’).

Todo

out_feature takes a list of features.

Example

>>> image = np.arange(5*5).reshape(5, 5)

>>> sap.create_profiles(image, {'area': [5, 10]},
...                     (sap.MinTree, sap.MaxTree))
Profiles{'attribute': 'area',
 'filtering rule': 'direct',
 'image': -7518820387991786804,
 'name': 'unknow',
 'out feature': 'altitude',
 'profiles': [{'operation': 'thinning', 'threshold': 10},
              {'operation': 'thinning', 'threshold': 5},
              {'operation': 'copy feature altitude'},
              {'operation': 'thickening', 'threshold': 5},
              {'operation': 'thickening', 'threshold': 10}]}

sap.profiles.differential(profiles)[source]¶

Compute the differential of profiles.

Parameters: profiles (Profiles) – Attribute profiles or other profiles to process the differential on.
Returns: differential – The processed differential profiles.
Return type: Profiles

sap.profiles.feature_profiles(image, attribute, adjacency=4, image_name=None, out_feature='same', filtering_rule='direct')[source]¶

Compute the feature profiles of an image.

Parameters

image (ndarray) – The image
attribute (dict) – Dictionary of attribute (as key, str) with according thresholds (as values, number).
adjacency (int) – Adjacency used for the tree construction. Default is 4.
image_name (str) – The name of the image (optional). Useful to track filtering process and display. If not set, the name is replaced by the hash of the image.
out_feature (str or list, optional) – Out feature of the profiles. Can be ‘altitude’ (default), ‘same’ or a list of feature. If ‘same’ then out feature of the profiles match the filtering attribute.
filtering_rule (str, optional) – The filtering rule to use. It can be ‘direct’, ‘min’, ‘max’ or ‘subtractive’. Default is ‘direct’.

Examples

>>> image = np.arange(5*5).reshape(5,5)

>>> sap.feature_profiles(image, {'area': [5, 10]})
Profiles{'attribute': 'area',
 'filtering rule': 'direct',
 'image': -7518820387991786804,
 'name': 'feature profiles',
 'out feature': 'area',
 'profiles': [{'operation': 'thinning', 'threshold': 10},
              {'operation': 'thinning', 'threshold': 5},
              {'operation': 'copy feature area'},
              {'operation': 'thickening', 'threshold': 5},
              {'operation': 'thickening', 'threshold': 10}]}

See also

sap.trees.available_attributes(): List available attributes.
attribute_profiles(): other profiles.

sap.profiles.local_features(profiles, local_feature=(np.mean, np.std), patch_size=7)[source]¶

Compute the local features of profiles

Parameters

profiles (Profiles) – Input Profiles.
local_feature (function or tuple of functions) – The function(s) to describe the local patches.
patch_size (int) – The size of the patches.

Returns

local_feature – The local features of profiles.

Return type

Profiles

sap.profiles.omega_profiles(image, attribute, adjacency=4, image_name=None, filtering_rule='direct')[source]¶

Compute the omega profiles of an image.

Parameters

image (ndarray) – The image
attribute (dict) – Dictionary of attribute (as key, str) with according thresholds (as values, number).
adjacency (int) – Adjacency used for the tree construction. Default is 4.
image_name (str) – The name of the image (optional). Useful to track filtering process and display. If not set, the name is replaced by the hash of the image.
filtering_rule (str, optional) – The filtering rule to use. It can be ‘direct’, ‘min’, ‘max’ or ‘subtractive’. Default is ‘direct’.

Examples

>>> image = np.arange(5 * 5).reshape(5, 5)
>>> sap.omega_profiles(image, {'area': [10, 100]})
Profiles{'attribute': 'area',
 'filtering rule': 'direct',
 'name': 'omega profiles',
 'out feature': 'altitude',
 'profiles': [{'operation': 'copy feature altitude'},
              {'operation': '(ω) filtering', 'threshold': 10},
              {'operation': '(ω) filtering', 'threshold': 100}],
 'tree': {'adjacency': 4, 'image_hash': '44f17c0f', 'image_name': None}}

See also

sap.trees.available_attributes(): List available attributes.

sap.profiles.self_dual_attribute_profiles(image, attribute, adjacency=4, image_name=None, filtering_rule='direct')[source]¶

Compute the self dual attribute profiles of an image.

Parameters

image (ndarray) – The image
attribute (dict) – Dictionary of attribute (as key, str) with according thresholds (as values, number).
adjacency (int) – Adjacency used for the tree construction. Default is 4.
image_name (str) – The name of the image (optional). Useful to track filtering process and display. If not set, the name is replaced by the hash of the image.
filtering_rule (str, optional) – The filtering rule to use. It can be ‘direct’, ‘min’, ‘max’ or ‘subtractive’. Default is ‘direct’.

Examples

>>> image = np.arange(5*5).reshape(5,5)

>>> sap.self_dual_attribute_profiles(image, {'area': [10, 100]})
Profiles{'attribute': 'area',
 'filtering rule': 'direct',
 'image': -7518820387991786804,
 'name': 'self dual attribute profiles',
 'out feature': 'altitude',
 'profiles': [{'operation': 'copy feature altitude'},
              {'operation': 'sd filtering', 'threshold': 10},
              {'operation': 'sd filtering', 'threshold': 100}]}

See also

sap.trees.available_attributes(): List available attributes.
attribute_profiles(): other profiles.

sap.profiles.self_dual_feature_profiles(image, attribute, adjacency=4, image_name=None, out_feature='same', filtering_rule='direct')[source]¶

Compute the self dual features profiles of an image.

Parameters

image (ndarray) – The image
attribute (dict) – Dictionary of attribute (as key, str) with according thresholds (as values, number).
adjacency (int) – Adjacency used for the tree construction. Default is 4.
image_name (str) – The name of the image (optional). Useful to track filtering process and display. If not set, the name is replaced by the hash of the image.
out_feature (str or list, optional) – Out feature of the profiles. Can be ‘altitude’ (default), ‘same’ or a list of feature. If ‘same’ then out feature of the profiles match the filtering attribute.
filtering_rule (str, optional) – The filtering rule to use. It can be ‘direct’, ‘min’, ‘max’ or ‘subtractive’. Default is ‘direct’.

Examples

>>> image = np.arange(5*5).reshape(5,5)

>>> sap.self_dual_feature_profiles(image, {'area': [10, 100]})
Profiles{'attribute': 'area',
 'filtering rule': 'direct',
 'image': -7518820387991786804,
 'name': 'self dual feature profiles',
 'out feature': 'area',
 'profiles': [{'operation': 'copy feature area'},
              {'operation': 'sd filtering', 'threshold': 10},
              {'operation': 'sd filtering', 'threshold': 100}]}

See also

sap.trees.available_attributes(): List available attributes.
attribute_profiles(): other profiles.

sap.profiles.show_all_profiles(profiles, attribute=None, image=None, height=None, fname=None, **kwargs)[source]¶

Display profiles with matplotlib.

Parameters

profiles (sap.profiles.Profiles) – The profiles to display.
attribute (sring, optional) – Name of attribute to display. By default display all the attributes contained in profiles.
image (string, optional) – Name of the image to display. By default display the profiles of all images.
height (scalar, optional, default: None) – Height of the figure in inches. Automatically adjust the size of the figure to display correctly the profiles and the title with matplot.
fname (str or PathLike, optional) – If set, the file path to save the figure. The attribute name is automatically inserted in the file name.

See also

show_profiles(): Display a profiles stack.

Notes

This is a utility function to call recursively show_profiles. Attribute and image filters are available to filter the profiles to display.

sap.profiles.show_profiles(profiles, height=None, fname=None, **kwargs)[source]¶

Display a profiles stack with matplotlib.

Parameters

profiles (Profiles) – The profiles to display. Can be only of length 1.
height (scalar, optional, default: None) – Height of the figure in inches. Automatically adjust the size of the figure to display correctly the profiles and the title with matplot.
fname (str or PathLike, optional) – If set, the file path to save the figure. The attribute name is automatically inserted in the file name.

See also

show_profiles_all(): Display several profiles at once.

sap.profiles.strip_profiles(lambda x: x['operation'] != 'thinning', profiles)[source]¶

Remove profiles according to condition. Iteration is done on profiles description (see Notes).

Parameters

condition (function) – The function (or lambda function) to use on profiles description to filter the profiles.
profiles (Profiles) – The profiles to filter.

Returns

new_profiles – Filtered profiles.

Return type

Profiles

Notes

The condition is tested on the description of each profiles. Considering this stack:

>>> aps
Profiles{'attribute': 'area',
 'image': -8884649894275650052,
 'profiles': [{'operation': 'thinning', 'threshold': 1000},
              {'operation': 'thinning', 'threshold': 100},
              {'operation': 'thinning', 'threshold': 10},
              {'operation': 'copy'},
              {'operation': 'thickening', 'threshold': 10},
              {'operation': 'thickening', 'threshold': 100},
              {'operation': 'thickening', 'threshold': 1000}]}

The condition function is tested on each item of the list 'profiles'.

See also

Profiles.strip(): Remove profiles based on condition.

Examples

Strip profiles depending on thresholds level:

>>> image = np.random.random((100, 100))
>>> aps = sap.attribute_profiles(image, {'area': [10, 100, 1000]})

>>> sap.strip_profiles(lambda x: 'threshold' in x and x['threshold'] > 20, aps)
Profiles{'attribute': 'area',
 'image': 2376333419322655105,
 'profiles': [{'operation': 'thinning', 'threshold': 10},
              {'operation': 'copy'},
              {'operation': 'thickening', 'threshold': 10}]}

Strip profiles depending on operation:

>>> sap.strip_profiles(lambda x: x['operation'] == 'thinning', aps)
Profiles{'attribute': 'area',
 'image': 2376333419322655105,
 'profiles': [{'operation': 'copy'},
              {'operation': 'thickening', 'threshold': 10},
              {'operation': 'thickening', 'threshold': 100},
              {'operation': 'thickening', 'threshold': 1000}]}

sap.profiles.strip_profiles_copy(profiles)[source]¶

Remove all the copied images in profiles.

Copy are the original images where profiles are computed on.

Parameters: profiles (Profiles) – The profiles to strip on the copied images.
Returns: new_profiles – Copy of profiles without copied image.
Return type: Profiles

See also

sap.strip_profiles(): Filter profiles according to condition.

sap.profiles.vectorize(profiles)[source]¶

Return the classification vectors of the profiles.

Parameters: profiles (Profiles) – Profiles on which process the vectors.
Returns: vectors – The vectors of the profiles.
Return type: numpy.ndarray

See also

Profiles.vectorize(): get the vectors of profiles.

Example

>>> image = np.random.random((100, 100))
>>> aps = sap.attribute_profiles(image, {'area': [10, 100]})

>>> vectors = sap.vectorize(aps)
>>> vectors.shape
(5, 100, 100)

Profiles¶

Simple Attribute Profiles

Navigation

Related Topics