用 python skimage做图像处理

导言：由于经常需要处理图像，习惯了matlab那么好用的图像处理工具箱，搬到python后有点不知所措了。搜了下python图像处理方面的库，好多PIL的教程，但是无奈的是PIL和numpy等结合不好，而且半天找不到图像处理的函数在哪里，只好作罢；终于搜到了skimage，:-)，喜出望外，因为它是和scipy、numpy可以完美结合的，那么就可以很好的使用numpy了。废话不多说，直接上干货吧。

step 1 概览

学习一个python库第一步是干啥？百度、Google搜教程？No，这不是我的风格。第一步就是直接 import。

import skimage
help(skimage)

看看输出：

Help on package skimage:

NAME
skimage - Image Processing SciKit (Toolbox for SciPy)

FILE
/usr/local/lib/python2.7/dist-packages/skimage/__init__.py

DESCRIPTION
``scikit-image`` (a.k.a. ``skimage``) is a collection of algorithms for image
processing and computer vision.

The main package of ``skimage`` only provides a few utilities for converting
between image data types; for most features, you need to import one of the
following subpackages:

Subpackages
-----------
color
Color space conversion.
data
Test images and example data.
draw
Drawing primitives (lines, text, etc.) that operate on NumPy arrays.
exposure
Image intensity adjustment, e.g., histogram equalization, etc.
feature
Feature detection and extraction, e.g., texture analysis corners, etc.
filters
Sharpening, edge finding, rank filters, thresholding, etc.
graph
Graph-theoretic operations, e.g., shortest paths.
io
Reading, saving, and displaying images and video.
measure
Measurement of image properties, e.g., similarity and contours.
morphology
Morphological operations, e.g., opening or skeletonization.
novice
Simplified interface for teaching purposes.
restoration
Restoration algorithms, e.g., deconvolution algorithms, denoising, etc.
segmentation
Partitioning an image into multiple regions.
transform
Geometric and other transforms, e.g., rotation or the Radon transform.
util
Generic utilities.
viewer
A simple graphical user interface for visualizing results and exploring
parameters.

Utility Functions
-----------------
img_as_float
Convert an image to floating point format, with values in [0, 1].
img_as_uint
Convert an image to unsigned integer format, with values in [0, 65535].
img_as_int
Convert an image to signed integer format, with values in [-32768, 32767].
img_as_ubyte
Convert an image to unsigned byte format, with values in [0, 255].

PACKAGE CONTENTS
_build
_shared (package)
color (package)
data (package)
draw (package)
exposure (package)
external (package)
feature (package)
filter (package)
filters (package)
future (package)
graph (package)
io (package)
measure (package)
morphology (package)
novice (package)
restoration (package)
scripts (package)
segmentation (package)
setup
transform (package)
util (package)
viewer (package)

FUNCTIONS
test = _test(doctest=False, verbose=False)
Run all unit tests.

DATA
__SKIMAGE_SETUP__ = False
__version__ = '0.12.3'
data_dir = '/usr/local/lib/python2.7/dist-packages/skimage/data'
doctest = <functools.partial object>
doctest_verbose = <functools.partial object>
pkg_dir = '/usr/local/lib/python2.7/dist-packages/skimage'
test_verbose = <functools.partial object>

VERSION
0.12.3

我们可以看到这里有好几个package是单独的，这意味着使用时要单独import。里面有那么多的库，每一个干啥用的？不要问，自己用help看看就知道了。

step 2 IO

我们这里看看如何读写图像文件。这个在skimage.io里面。

import skimage, skimage.io
help(skimage.io)

Help on package skimage.io in skimage:

NAME
skimage.io - Utilities to read and write images in various formats.

FILE
/usr/local/lib/python2.7/dist-packages/skimage/io/__init__.py

DESCRIPTION
The following plug-ins are available:

========== ==============================================================
Plugin     Description
---------- --------------------------------------------------------------
pil        Image reading via the Python Imaging Library
qt         Fast image display using the Qt library
freeimage  Load images using the FreeImage library
gtk        Fast image display using the GTK library
matplotlib Display or save images using Matplotlib
fits       FITS image reading via PyFITS
simpleitk  Image reading and writing via SimpleITK
imageio    Image reading via the ImageIO Library
imread     Image reading and writing via imread
tifffile   Load and save TIFF and TIFF-based images using tifffile.py
gdal       Image reading via the GDAL Library (www.gdal.org)
========== ==============================================================

PACKAGE CONTENTS
_image_stack
_io
_plugins (package)
collection
manage_plugins
setup
sift
tests (package)
util

DATA
WRAP_LEN = 73
available_plugins = {'fits': ['imread', 'imread_collection'], 'freeima...
image_stack = []

好像没有列出全部的函数，没关系，继续看。

dir(skimage.io)

['ImageCollection',
'MultiImage',
'WRAP_LEN',
'__builtins__',
'__doc__',
'__file__',
'__name__',
'__package__',
'__path__',
'_format_plugin_info_table',
'_image_stack',
'_io',
'_plugins',
'_separator',
'_update_doc',
'available_plugins',
'call_plugin',
'collection',
'concatenate_images',
'find_available_plugins',
'image_stack',
'imread',
'imread_collection',
'imread_collection_wrapper',
'imsave',
'imshow',
'imshow_collection',
'load_sift',
'load_surf',
'manage_plugins',
'plugin_info',
'plugin_order',
'pop',
'push',
'reset_plugins',
'show',
'sift',
'use_plugin',
'util']

imread： 和matlab一样，就是读取图像

imsave: 和matlab一样，就是存储图像

imshow: 调用的matplotlib的显示，所以必须再使用show才能真正显示。

其他的就不多说了，自己help就很清楚了。

step 3 形态学变换

由step1可以看出形态学变换的东西肯定就在 skimage.morphology里面了。

import skimage, skimage.morphology
help(skimage.morphology)

Help on package skimage.morphology in skimage:

NAME
skimage.morphology

FILE
/usr/local/lib/python2.7/dist-packages/skimage/morphology/__init__.py

PACKAGE CONTENTS
_convex_hull
_greyreconstruct
_skeletonize
_skeletonize_3d
_skeletonize_3d_cy
_skeletonize_cy
_watershed
binary
convex_hull
grey
greyreconstruct
misc
selem
setup
tests (package)
watershed

FUNCTIONS
ball(radius, dtype=<type 'numpy.uint8'>)
Generates a ball-shaped structuring element.

This is the 3D equivalent of a disk.
A pixel is within the neighborhood if the euclidean distance between
it and the origin is no greater than radius.

Parameters
----------
radius : int
The radius of the ball-shaped structuring element.

Other Parameters
----------------
dtype : data-type
The data type of the structuring element.

Returns
-------
selem : ndarray
The structuring element where elements of the neighborhood
are 1 and 0 otherwise.

binary_closing(image, selem=None, *args, **kwargs)
Return fast binary morphological closing of an image.

This function returns the same result as greyscale closing but performs
faster for binary images.

The morphological closing on an image is defined as a dilation followed by
an erosion. Closing can remove small dark spots (i.e. "pepper") and connect
small bright cracks. This tends to "close" up (dark) gaps between (bright)
features.

Parameters
----------
image : ndarray
Binary input image.
selem : ndarray, optional
The neighborhood expressed as a 2-D array of 1's and 0's.
If None, use cross-shaped structuring element (connectivity=1).
out : ndarray of bool, optional
The array to store the result of the morphology. If None,
is passed, a new array will be allocated.

Returns
-------
closing : ndarray of bool
The result of the morphological closing.

binary_dilation(image, selem=None, *args, **kwargs)
Return fast binary morphological dilation of an image.

This function returns the same result as greyscale dilation but performs
faster for binary images.

Morphological dilation sets a pixel at ``(i,j)`` to the maximum over all
pixels in the neighborhood centered at ``(i,j)``. Dilation enlarges bright
regions and shrinks dark regions.

Parameters
----------

image : ndarray
Binary input image.
selem : ndarray, optional
The neighborhood expressed as a 2-D array of 1's and 0's.
If None, use cross-shaped structuring element (connectivity=1).
out : ndarray of bool, optional
The array to store the result of the morphology. If None, is
passed, a new array will be allocated.

Returns
-------
dilated : ndarray of bool or uint
The result of the morphological dilation with values in
``[False, True]``.

binary_erosion(image, selem=None, *args, **kwargs)
Return fast binary morphological erosion of an image.

This function returns the same result as greyscale erosion but performs
faster for binary images.

Morphological erosion sets a pixel at ``(i,j)`` to the minimum over all
pixels in the neighborhood centered at ``(i,j)``. Erosion shrinks bright
regions and enlarges dark regions.

Parameters
----------
image : ndarray
Binary input image.
selem : ndarray, optional
The neighborhood expressed as a 2-D array of 1's and 0's.
If None, use cross-shaped structuring element (connectivity=1).
out : ndarray of bool, optional
The array to store the result of the morphology. If None is
passed, a new array will be allocated.

Returns
-------
eroded : ndarray of bool or uint
The result of the morphological erosion taking values in
``[False, True]``.

binary_opening(image, selem=None, *args, **kwargs)
Return fast binary morphological opening of an image.

This function returns the same result as greyscale opening but performs
faster for binary images.

The morphological opening on an image is defined as an erosion followed by
a dilation. Opening can remove small bright spots (i.e. "salt") and connect
small dark cracks. This tends to "open" up (dark) gaps between (bright)
features.

Parameters
----------
image : ndarray
Binary input image.
selem : ndarray, optional
The neighborhood expressed as a 2-D array of 1's and 0's.
If None, use cross-shaped structuring element (connectivity=1).
out : ndarray of bool, optional
The array to store the result of the morphology. If None
is passed, a new array will be allocated.

Returns
-------
opening : ndarray of bool
The result of the morphological opening.

black_tophat(image, selem=None, *args, **kwargs)
Return black top hat of an image.

The black top hat of an image is defined as its morphological closing minus
the original image. This operation returns the dark spots of the image that
are smaller than the structuring element. Note that dark spots in the
original image are bright spots after the black top hat.

Parameters
----------
image : ndarray
Image array.
selem : ndarray, optional
The neighborhood expressed as a 2-D array of 1's and 0's.
If None, use cross-shaped structuring element (connectivity=1).
out : ndarray, optional
The array to store the result of the morphology. If None
is passed, a new array will be allocated.

Returns
-------
opening : array, same shape and type as `image`
The result of the black top filter.

Examples
--------
>>> # Change dark peak to bright peak and subtract background
>>> import numpy as np
>>> from skimage.morphology import square
>>> dark_on_grey = np.array([[7, 6, 6, 6, 7],
...                          [6, 5, 4, 5, 6],
...                          [6, 4, 0, 4, 6],
...                          [6, 5, 4, 5, 6],
...                          [7, 6, 6, 6, 7]], dtype=np.uint8)
>>> black_tophat(dark_on_grey, square(3))
array([[0, 0, 0, 0, 0],
[0, 0, 1, 0, 0],
[0, 1, 5, 1, 0],
[0, 0, 1, 0, 0],
[0, 0, 0, 0, 0]], dtype=uint8)

closing(image, selem=None, *args, **kwargs)
Return greyscale morphological closing of an image.

The morphological closing on an image is defined as a dilation followed by
an erosion. Closing can remove small dark spots (i.e. "pepper") and connect
small bright cracks. This tends to "close" up (dark) gaps between (bright)
features.

Parameters
----------
image : ndarray
Image array.
selem : ndarray, optional
The neighborhood expressed as an array of 1's and 0's.
If None, use cross-shaped structuring element (connectivity=1).
out : ndarray, optional
The array to store the result of the morphology. If None,
is passed, a new array will be allocated.

Returns
-------
closing : array, same shape and type as `image`
The result of the morphological closing.

Examples
--------
>>> # Close a gap between two bright lines
>>> import numpy as np
>>> from skimage.morphology import square
>>> broken_line = np.array([[0, 0, 0, 0, 0],
...                         [0, 0, 0, 0, 0],
...                         [1, 1, 0, 1, 1],
...                         [0, 0, 0, 0, 0],
...                         [0, 0, 0, 0, 0]], dtype=np.uint8)
>>> closing(broken_line, square(3))
array([[0, 0, 0, 0, 0],
[0, 0, 0, 0, 0],
[1, 1, 1, 1, 1],
[0, 0, 0, 0, 0],
[0, 0, 0, 0, 0]], dtype=uint8)

convex_hull_image(image)
Compute the convex hull image of a binary image.

The convex hull is the set of pixels included in the smallest convex
polygon that surround all white pixels in the input image.

Parameters
----------
image : (M, N) array
Binary input image. This array is cast to bool before processing.

Returns
-------
hull : (M, N) array of bool
Binary image with pixels in convex hull set to True.

References
----------
.. [1] http://blogs.mathworks.com/steve/2011/10/04/binary-image-convex-hull-algorithm-notes/ 
convex_hull_object(image, neighbors=8)
Compute the convex hull image of individual objects in a binary image.

The convex hull is the set of pixels included in the smallest convex
polygon that surround all white pixels in the input image.

Parameters
----------
image : (M, N) array
Binary input image.
neighbors : {4, 8}, int
Whether to use 4- or 8-connectivity.

Returns
-------
hull : ndarray of bool
Binary image with pixels in convex hull set to True.

Notes
-----
This function uses skimage.morphology.label to define unique objects,
finds the convex hull of each using convex_hull_image, and combines
these regions with logical OR. Be aware the convex hulls of unconnected
objects may overlap in the result. If this is suspected, consider using
convex_hull_image separately on each object.

cube(width, dtype=<type 'numpy.uint8'>)
Generates a cube-shaped structuring element.

This is the 3D equivalent of a square.
Every pixel along the perimeter has a chessboard distance
no greater than radius (radius=floor(width/2)) pixels.

Parameters
----------
width : int
The width, height and depth of the cube.

Other Parameters
----------------
dtype : data-type
The data type of the structuring element.

Returns
-------
selem : ndarray
A structuring element consisting only of ones, i.e. every
pixel belongs to the neighborhood.

diamond(radius, dtype=<type 'numpy.uint8'>)
Generates a flat, diamond-shaped structuring element.

A pixel is part of the neighborhood (i.e. labeled 1) if
the city block/Manhattan distance between it and the center of
the neighborhood is no greater than radius.

Parameters
----------
radius : int
The radius of the diamond-shaped structuring element.

Other Parameters
----------------
dtype : data-type
The data type of the structuring element.

Returns
-------

selem : ndarray
The structuring element where elements of the neighborhood
are 1 and 0 otherwise.

dilation(image, selem=None, *args, **kwargs)
Return greyscale morphological dilation of an image.

Morphological dilation sets a pixel at (i,j) to the maximum over all pixels
in the neighborhood centered at (i,j). Dilation enlarges bright regions
and shrinks dark regions.

Parameters
----------

image : ndarray
Image array.
selem : ndarray, optional
The neighborhood expressed as a 2-D array of 1's and 0's.
If None, use cross-shaped structuring element (connectivity=1).
out : ndarray, optional
The array to store the result of the morphology. If None, is
passed, a new array will be allocated.
shift_x, shift_y : bool, optional
shift structuring element about center point. This only affects
eccentric structuring elements (i.e. selem with even numbered sides).

Returns
-------
dilated : uint8 array, same shape and type as `image`
The result of the morphological dilation.

Notes
-----
For `uint8` (and `uint16` up to a certain bit-depth) data, the lower
algorithm complexity makes the `skimage.filter.rank.maximum` function more
efficient for larger images and structuring elements.

Examples
--------
>>> # Dilation enlarges bright regions
>>> import numpy as np
>>> from skimage.morphology import square
>>> bright_pixel = np.array([[0, 0, 0, 0, 0],
...                          [0, 0, 0, 0, 0],
...                          [0, 0, 1, 0, 0],
...                          [0, 0, 0, 0, 0],
...                          [0, 0, 0, 0, 0]], dtype=np.uint8)
>>> dilation(bright_pixel, square(3))
array([[0, 0, 0, 0, 0],
[0, 1, 1, 1, 0],
[0, 1, 1, 1, 0],
[0, 1, 1, 1, 0],
[0, 0, 0, 0, 0]], dtype=uint8)

disk(radius, dtype=<type 'numpy.uint8'>)
Generates a flat, disk-shaped structuring element.

A pixel is within the neighborhood if the euclidean distance between
it and the origin is no greater than radius.

Parameters
----------
radius : int
The radius of the disk-shaped structuring element.

Other Parameters
----------------
dtype : data-type
The data type of the structuring element.

Returns
-------
selem : ndarray
The structuring element where elements of the neighborhood
are 1 and 0 otherwise.

erosion(image, selem=None, *args, **kwargs)
Return greyscale morphological erosion of an image.

Morphological erosion sets a pixel at (i,j) to the minimum over all pixels
in the neighborhood centered at (i,j). Erosion shrinks bright regions and
enlarges dark regions.

Parameters
----------
image : ndarray
Image array.
selem : ndarray, optional
The neighborhood expressed as an array of 1's and 0's.
If None, use cross-shaped structuring element (connectivity=1).
out : ndarrays, optional
The array to store the result of the morphology. If None is
passed, a new array will be allocated.
shift_x, shift_y : bool, optional
shift structuring element about center point. This only affects
eccentric structuring elements (i.e. selem with even numbered sides).

Returns
-------
eroded : array, same shape as `image`
The result of the morphological erosion.

Notes
-----
For ``uint8`` (and ``uint16`` up to a certain bit-depth) data, the
lower algorithm complexity makes the `skimage.filter.rank.minimum`
function more efficient for larger images and structuring elements.

Examples
--------
>>> # Erosion shrinks bright regions
>>> import numpy as np
>>> from skimage.morphology import square
>>> bright_square = np.array([[0, 0, 0, 0, 0],
...                           [0, 1, 1, 1, 0],
...                           [0, 1, 1, 1, 0],
...                           [0, 1, 1, 1, 0],
...                           [0, 0, 0, 0, 0]], dtype=np.uint8)
>>> erosion(bright_square, square(3))
array([[0, 0, 0, 0, 0],
[0, 0, 0, 0, 0],
[0, 0, 1, 0, 0],
[0, 0, 0, 0, 0],
[0, 0, 0, 0, 0]], dtype=uint8)

label(input, neighbors=None, background=None, return_num=False, connectivity=None)
Label connected regions of an integer array.

Two pixels are connected when they are neighbors and have the same value.
In 2D, they can be neighbors either in a 1- or 2-connected sense.
The value refers to the maximum number of orthogonal hops to consider a
pixel/voxel a neighbor::

1-connectivity      2-connectivity     diagonal connection close-up

[ ]           [ ]  [ ]  [ ]         [ ]
|               \  |  /             |  <- hop 2
[ ]--[x]--[ ]      [ ]--[x]--[ ]    [x]--[ ]
|               /  |  \         hop 1
[ ]           [ ]  [ ]  [ ]

Parameters
----------
input : ndarray of dtype int
Image to label.
neighbors : {4, 8}, int, optional
Whether to use 4- or 8-"connectivity".
In 3D, 4-"connectivity" means connected pixels have to share face,
whereas with 8-"connectivity", they have to share only edge or vertex.
**Deprecated, use ``connectivity`` instead.**
background : int, optional
Consider all pixels with this value as background pixels, and label
them as 0. By default, 0-valued pixels are considered as background
pixels.
return_num : bool, optional
Whether to return the number of assigned labels.
connectivity : int, optional
Maximum number of orthogonal hops to consider a pixel/voxel
as a neighbor.
Accepted values are ranging from  1 to input.ndim. If ``None``, a full
connectivity of ``input.ndim`` is used.

Returns
-------
labels : ndarray of dtype int
Labeled array, where all connected regions are assigned the
same integer value.
num : int, optional
Number of labels, which equals the maximum label index and is only
returned if return_num is `True`.

Examples
--------
>>> import numpy as np
>>> x = np.eye(3).astype(int)
>>> print(x)
[[1 0 0]
[0 1 0]
[0 0 1]]
>>> from skimage.measure import label
>>> print(label(x, connectivity=1))
[[1 0 0]
[0 2 0]
[0 0 3]]

>>> print(label(x, connectivity=2))
[[1 0 0]
[0 1 0]
[0 0 1]]

>>> print(label(x, background=-1))
[[1 2 2]
[2 1 2]
[2 2 1]]

>>> x = np.array([[1, 0, 0],
...               [1, 1, 5],
...               [0, 0, 0]])

>>> print(label(x))
[[1 0 0]
[1 1 2]
[0 0 0]]

medial_axis(image, mask=None, return_distance=False)
Compute the medial axis transform of a binary image

Parameters
----------
image : binary ndarray, shape (M, N)
The image of the shape to be skeletonized.
mask : binary ndarray, shape (M, N), optional
If a mask is given, only those elements in `image` with a true
value in `mask` are used for computing the medial axis.
return_distance : bool, optional
If true, the distance transform is returned as well as the skeleton.

Returns
-------
out : ndarray of bools
Medial axis transform of the image
dist : ndarray of ints, optional
Distance transform of the image (only returned if `return_distance`
is True)

See also
--------
skeletonize

Notes
-----
This algorithm computes the medial axis transform of an image
as the ridges of its distance transform.

The different steps of the algorithm are as follows
* A lookup table is used, that assigns 0 or 1 to each configuration of
the 3x3 binary square, whether the central pixel should be removed
or kept. We want a point to be removed if it has more than one neighbor
and if removing it does not change the number of connected components.

* The distance transform to the background is computed, as well as
the cornerness of the pixel.

* The foreground (value of 1) points are ordered by
the distance transform, then the cornerness.

* A cython function is called to reduce the image to its skeleton. It
processes pixels in the order determined at the previous step, and
removes or maintains a pixel according to the lookup table. Because
of the ordering, it is possible to process all pixels in only one
pass.

Examples
--------
>>> square = np.zeros((7, 7), dtype=np.uint8)
>>> square[1:-1, 2:-2] = 1
>>> square
array([[0, 0, 0, 0, 0, 0, 0],
[0, 0, 1, 1, 1, 0, 0],
[0, 0, 1, 1, 1, 0, 0],
[0, 0, 1, 1, 1, 0, 0],
[0, 0, 1, 1, 1, 0, 0],
[0, 0, 1, 1, 1, 0, 0],
[0, 0, 0, 0, 0, 0, 0]], dtype=uint8)
>>> medial_axis(square).astype(np.uint8)
array([[0, 0, 0, 0, 0, 0, 0],
[0, 0, 1, 0, 1, 0, 0],
[0, 0, 0, 1, 0, 0, 0],
[0, 0, 0, 1, 0, 0, 0],
[0, 0, 0, 1, 0, 0, 0],
[0, 0, 1, 0, 1, 0, 0],
[0, 0, 0, 0, 0, 0, 0]], dtype=uint8)

octagon(m, n, dtype=<type 'numpy.uint8'>)
Generates an octagon shaped structuring element.

For a given size of (m) horizontal and vertical sides
and a given (n) height or width of slanted sides octagon is generated.
The slanted sides are 45 or 135 degrees to the horizontal axis
and hence the widths and heights are equal.

Parameters
----------
m : int
The size of the horizontal and vertical sides.
n : int
The height or width of the slanted sides.

Other Parameters
----------------
dtype : data-type
The data type of the structuring element.

Returns
-------
selem : ndarray
The structuring element where elements of the neighborhood
are 1 and 0 otherwise.

octahedron(radius, dtype=<type 'numpy.uint8'>)
Generates a octahedron-shaped structuring element.

This is the 3D equivalent of a diamond.
A pixel is part of the neighborhood (i.e. labeled 1) if
the city block/Manhattan distance between it and the center of
the neighborhood is no greater than radius.

Parameters
----------
radius : int
The radius of the octahedron-shaped structuring element.

Other Parameters
----------------
dtype : data-type
The data type of the structuring element.

Returns
-------

selem : ndarray
The structuring element where elements of the neighborhood
are 1 and 0 otherwise.

opening(image, selem=None, *args, **kwargs)
Return greyscale morphological opening of an image.

The morphological opening on an image is defined as an erosion followed by
a dilation. Opening can remove small bright spots (i.e. "salt") and connect
small dark cracks. This tends to "open" up (dark) gaps between (bright)
features.

Parameters
----------
image : ndarray
Image array.
selem : ndarray, optional
The neighborhood expressed as an array of 1's and 0's.
If None, use cross-shaped structuring element (connectivity=1).
out : ndarray, optional
The array to store the result of the morphology. If None
is passed, a new array will be allocated.

Returns
-------
opening : array, same shape and type as `image`
The result of the morphological opening.

Examples
--------
>>> # Open up gap between two bright regions (but also shrink regions)
>>> import numpy as np
>>> from skimage.morphology import square
>>> bad_connection = np.array([[1, 0, 0, 0, 1],
...                            [1, 1, 0, 1, 1],
...                            [1, 1, 1, 1, 1],
...                            [1, 1, 0, 1, 1],
...                            [1, 0, 0, 0, 1]], dtype=np.uint8)
>>> opening(bad_connection, square(3))
array([[0, 0, 0, 0, 0],
[1, 1, 0, 1, 1],
[1, 1, 0, 1, 1],
[1, 1, 0, 1, 1],
[0, 0, 0, 0, 0]], dtype=uint8)

reconstruction(seed, mask, method='dilation', selem=None, offset=None)
Perform a morphological reconstruction of an image.

Morphological reconstruction by dilation is similar to basic morphological
dilation: high-intensity values will replace nearby low-intensity values.
The basic dilation operator, however, uses a structuring element to
determine how far a value in the input image can spread. In contrast,
reconstruction uses two images: a "seed" image, which specifies the values
that spread, and a "mask" image, which gives the maximum allowed value at
each pixel. The mask image, like the structuring element, limits the spread
of high-intensity values. Reconstruction by erosion is simply the inverse:
low-intensity values spread from the seed image and are limited by the mask
image, which represents the minimum allowed value.

Alternatively, you can think of reconstruction as a way to isolate the
connected regions of an image. For dilation, reconstruction connects
regions marked by local maxima in the seed image: neighboring pixels
less-than-or-equal-to those seeds are connected to the seeded region.
Local maxima with values larger than the seed image will get truncated to
the seed value.

Parameters
----------
seed : ndarray
The seed image (a.k.a. marker image), which specifies the values that
are dilated or eroded.
mask : ndarray
The maximum (dilation) / minimum (erosion) allowed value at each pixel.
method : {'dilation'|'erosion'}
Perform reconstruction by dilation or erosion. In dilation (or
erosion), the seed image is dilated (or eroded) until limited by the
mask image. For dilation, each seed value must be less than or equal
to the corresponding mask value; for erosion, the reverse is true.
selem : ndarray
The neighborhood expressed as a 2-D array of 1's and 0's.

Returns
-------
reconstructed : ndarray
The result of morphological reconstruction.

Examples
--------
>>> import numpy as np
>>> from skimage.morphology import reconstruction

First, we create a sinusoidal mask image with peaks at middle and ends.

>>> x = np.linspace(0, 4 * np.pi)
>>> y_mask = np.cos(x)

Then, we create a seed image initialized to the minimum mask value (for
reconstruction by dilation, min-intensity values don't spread) and add
"seeds" to the left and right peak, but at a fraction of peak value (1).

>>> y_seed = y_mask.min() * np.ones_like(x)
>>> y_seed[0] = 0.5
>>> y_seed[-1] = 0
>>> y_rec = reconstruction(y_seed, y_mask)

The reconstructed image (or curve, in this case) is exactly the same as the
mask image, except that the peaks are truncated to 0.5 and 0. The middle
peak disappears completely: Since there were no seed values in this peak
region, its reconstructed value is truncated to the surrounding value (-1).

As a more practical example, we try to extract the bright features of an
image by subtracting a background image created by reconstruction.

>>> y, x = np.mgrid[:20:0.5, :20:0.5]
>>> bumps = np.sin(x) + np.sin(y)

To create the background image, set the mask image to the original image,
and the seed image to the original image with an intensity offset, `h`.

>>> h = 0.3
>>> seed = bumps - h
>>> background = reconstruction(seed, bumps)

The resulting reconstructed image looks exactly like the original image,
but with the peaks of the bumps cut off. Subtracting this reconstructed
image from the original image leaves just the peaks of the bumps

>>> hdome = bumps - background

This operation is known as the h-dome of the image and leaves features
of height `h` in the subtracted image.

Notes
-----
The algorithm is taken from [1]_. Applications for greyscale reconstruction
are discussed in [2]_ and [3]_.

References
----------
.. [1] Robinson, "Efficient morphological reconstruction: a downhill
filter", Pattern Recognition Letters 25 (2004) 1759-1767.
.. [2] Vincent, L., "Morphological Grayscale Reconstruction in Image
Analysis: Applications and Efficient Algorithms", IEEE Transactions
on Image Processing (1993)
.. [3] Soille, P., "Morphological Image Analysis: Principles and
Applications", Chapter 6, 2nd edition (2003), ISBN 3540429883.

rectangle(width, height, dtype=<type 'numpy.uint8'>)
Generates a flat, rectangular-shaped structuring element.

Every pixel in the rectangle generated for a given width and given height
belongs to the neighborhood.

Parameters
----------
width : int
The width of the rectangle.
height : int
The height of the rectangle.

Other Parameters
----------------
dtype : data-type
The data type of the structuring element.

Returns
-------
selem : ndarray
A structuring element consisting only of ones, i.e. every
pixel belongs to the neighborhood.

remove_small_holes(ar, min_size=64, connectivity=1, in_place=False)
Remove continguous holes smaller than the specified size.

Parameters
----------
ar : ndarray (arbitrary shape, int or bool type)
The array containing the connected components of interest.
min_size : int, optional (default: 64)
The hole component size.
connectivity : int, {1, 2, ..., ar.ndim}, optional (default: 1)
The connectivity defining the neighborhood of a pixel.
in_place : bool, optional (default: False)
If `True`, remove the connected components in the input array itself.
Otherwise, make a copy.

Raises
------
TypeError
If the input array is of an invalid type, such as float or string.
ValueError
If the input array contains negative values.

Returns
-------
out : ndarray, same shape and type as input `ar`
The input array with small holes within connected components removed.

Examples
--------
>>> from skimage import morphology
>>> a = np.array([[1, 1, 1, 1, 1, 0],
...               [1, 1, 1, 0, 1, 0],
...               [1, 0, 0, 1, 1, 0],
...               [1, 1, 1, 1, 1, 0]], bool)
>>> b = morphology.remove_small_holes(a, 2)
>>> b
array([[ True,  True,  True,  True,  True, False],
[ True,  True,  True,  True,  True, False],
[ True, False, False,  True,  True, False],
[ True,  True,  True,  True,  True, False]], dtype=bool)
>>> c = morphology.remove_small_holes(a, 2, connectivity=2)
>>> c
array([[ True,  True,  True,  True,  True, False],
[ True,  True,  True, False,  True, False],
[ True, False, False,  True,  True, False],
[ True,  True,  True,  True,  True, False]], dtype=bool)
>>> d = morphology.remove_small_holes(a, 2, in_place=True)
>>> d is a
True

Notes
-----

If the array type is int, it is assumed that it contains already-labeled
objects. The labels are not kept in the output image (this function always
outputs a bool image). It is suggested that labeling is completed after
using this function.

remove_small_objects(ar, min_size=64, connectivity=1, in_place=False)
Remove connected components smaller than the specified size.

Parameters
----------
ar : ndarray (arbitrary shape, int or bool type)
The array containing the connected components of interest. If the array
type is int, it is assumed that it contains already-labeled objects.
The ints must be non-negative.
min_size : int, optional (default: 64)
The smallest allowable connected component size.
connectivity : int, {1, 2, ..., ar.ndim}, optional (default: 1)
The connectivity defining the neighborhood of a pixel.
in_place : bool, optional (default: False)
If `True`, remove the connected components in the input array itself.
Otherwise, make a copy.

Raises
------
TypeError
If the input array is of an invalid type, such as float or string.
ValueError
If the input array contains negative values.

Returns
-------
out : ndarray, same shape and type as input `ar`
The input array with small connected components removed.

Examples
--------
>>> from skimage import morphology
>>> a = np.array([[0, 0, 0, 1, 0],
...               [1, 1, 1, 0, 0],
...               [1, 1, 1, 0, 1]], bool)
>>> b = morphology.remove_small_objects(a, 6)
>>> b
array([[False, False, False, False, False],
[ True,  True,  True, False, False],
[ True,  True,  True, False, False]], dtype=bool)
>>> c = morphology.remove_small_objects(a, 7, connectivity=2)
>>> c
array([[False, False, False,  True, False],
[ True,  True,  True, False, False],
[ True,  True,  True, False, False]], dtype=bool)
>>> d = morphology.remove_small_objects(a, 6, in_place=True)
>>> d is a
True

skeletonize(image)
Return the skeleton of a binary image.

Thinning is used to reduce each connected component in a binary image
to a single-pixel wide skeleton.

Parameters
----------
image : numpy.ndarray
A binary image containing the objects to be skeletonized. '1'
represents foreground, and '0' represents background. It
also accepts arrays of boolean values where True is foreground.

Returns
-------
skeleton : ndarray
A matrix containing the thinned image.

See also
--------
medial_axis

Notes
-----
The algorithm [1]_ works by making successive passes of the image,
removing pixels on object borders. This continues until no
more pixels can be removed.  The image is correlated with a
mask that assigns each pixel a number in the range [0...255]
corresponding to each possible pattern of its 8 neighbouring
pixels. A look up table is then used to assign the pixels a
value of 0, 1, 2 or 3, which are selectively removed during
the iterations.

Note that this algorithm will give different results than a
medial axis transform, which is also often referred to as
"skeletonization".

References
----------
.. [1] A fast parallel algorithm for thinning digital patterns,
T. Y. Zhang and C. Y. Suen, Communications of the ACM,
March 1984, Volume 27, Number 3.

Examples
--------
>>> X, Y = np.ogrid[0:9, 0:9]
>>> ellipse = (1./3 * (X - 4)**2 + (Y - 4)**2 < 3**2).astype(np.uint8)
>>> ellipse
array([[0, 0, 0, 1, 1, 1, 0, 0, 0],
[0, 0, 1, 1, 1, 1, 1, 0, 0],
[0, 0, 1, 1, 1, 1, 1, 0, 0],
[0, 0, 1, 1, 1, 1, 1, 0, 0],
[0, 0, 1, 1, 1, 1, 1, 0, 0],
[0, 0, 1, 1, 1, 1, 1, 0, 0],
[0, 0, 1, 1, 1, 1, 1, 0, 0],
[0, 0, 1, 1, 1, 1, 1, 0, 0],
[0, 0, 0, 1, 1, 1, 0, 0, 0]], dtype=uint8)
>>> skel = skeletonize(ellipse)
>>> skel.astype(np.uint8)
array([[0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 0, 0, 0, 0],
[0, 0, 0, 0, 1, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0]], dtype=uint8)

skeletonize_3d(img)
Compute the skeleton of a binary image.

Thinning is used to reduce each connected component in a binary image
to a single-pixel wide skeleton.

Parameters
----------
img : ndarray, 2D or 3D
A binary image containing the objects to be skeletonized. Zeros
represent background, nonzero values are foreground.

Returns
-------
skeleton : ndarray
The thinned image.

See also
--------
skeletonize, medial_axis

Notes
-----
The method of [Lee94]_ uses an octree data structure to examine a 3x3x3
neighborhood of a pixel. The algorithm proceeds by iteratively sweeping
over the image, and removing pixels at each iteration until the image
stops changing. Each iteration consists of two steps: first, a list of
candidates for removal is assembled; then pixels from this list are
rechecked sequentially, to better preserve connectivity of the image.

The algorithm this function implements is different from the algorithms
used by either `skeletonize` or `medial_axis`, thus for 2D images the
results produced by this function are generally different.

References
----------
.. [Lee94] T.-C. Lee, R.L. Kashyap and C.-N. Chu, Building skeleton models
via 3-D medial surface/axis thinning algorithms.
Computer Vision, Graphics, and Image Processing, 56(6):462-478, 1994.

square(width, dtype=<type 'numpy.uint8'>)
Generates a flat, square-shaped structuring element.

Every pixel along the perimeter has a chessboard distance
no greater than radius (radius=floor(width/2)) pixels.

Parameters
----------
width : int
The width and height of the square.

Other Parameters
----------------
dtype : data-type
The data type of the structuring element.

Returns
-------
selem : ndarray
A structuring element consisting only of ones, i.e. every
pixel belongs to the neighborhood.

watershed(image, markers, connectivity=None, offset=None, mask=None)
Return a matrix labeled using the watershed segmentation algorithm

Parameters
----------

image: ndarray (2-D, 3-D, ...) of integers
Data array where the lowest value points are labeled first.
markers: ndarray of the same shape as `image`
An array marking the basins with the values to be assigned in the
label matrix. Zero means not a marker. This array should be of an
integer type.
connectivity: ndarray, optional
An array with the same number of dimensions as `image` whose
non-zero elements indicate neighbors for connection.
Following the scipy convention, default is a one-connected array of
the dimension of the image.
offset: array_like of shape image.ndim, optional
offset of the connectivity (one offset per dimension)
mask: ndarray of bools or 0s and 1s, optional
Array of same shape as `image`. Only points at which mask == True
will be labeled.

Returns
-------
out: ndarray
A labeled matrix of the same type and shape as markers

See also
--------

skimage.segmentation.random_walker: random walker segmentation
A segmentation algorithm based on anisotropic diffusion, usually
slower than the watershed but with good results on noisy data and
boundaries with holes.

Notes
-----
This function implements a watershed algorithm [1]_that apportions pixels
into marked basins. The algorithm uses a priority queue to hold the pixels
with the metric for the priority queue being pixel value, then the time of
entry into the queue - this settles ties in favor of the closest marker.

Some ideas taken from
Soille, "Automated Basin Delineation from Digital Elevation Models Using
Mathematical Morphology", Signal Processing 20 (1990) 171-182

The most important insight in the paper is that entry time onto the queue
solves two problems: a pixel should be assigned to the neighbor with the
largest gradient or, if there is no gradient, pixels on a plateau should
be split between markers on opposite sides.

This implementation converts all arguments to specific, lowest common
denominator types, then passes these to a C algorithm.

Markers can be determined manually, or automatically using for example
the local minima of the gradient of the image, or the local maxima of the
distance function to the background for separating overlapping objects
(see example).

References
----------
.. [1] http://en.wikipedia.org/wiki/Watershed_%28image_processing%29 
.. [2] http://cmm.ensmp.fr/~beucher/wtshed.html 
Examples
--------
The watershed algorithm is very useful to separate overlapping objects

>>> # Generate an initial image with two overlapping circles
>>> x, y = np.indices((80, 80))
>>> x1, y1, x2, y2 = 28, 28, 44, 52
>>> r1, r2 = 16, 20
>>> mask_circle1 = (x - x1)**2 + (y - y1)**2 < r1**2
>>> mask_circle2 = (x - x2)**2 + (y - y2)**2 < r2**2
>>> image = np.logical_or(mask_circle1, mask_circle2)
>>> # Now we want to separate the two objects in image
>>> # Generate the markers as local maxima of the distance
>>> # to the background
>>> from scipy import ndimage as ndi
>>> distance = ndi.distance_transform_edt(image)
>>> from skimage.feature import peak_local_max
>>> local_maxi = peak_local_max(distance, labels=image,
...                             footprint=np.ones((3, 3)),
...                             indices=False)
>>> markers = ndi.label(local_maxi)[0]
>>> labels = watershed(-distance, markers, mask=image)

The algorithm works also for 3-D images, and can be used for example to
separate overlapping spheres.

white_tophat(image, selem=None, *args, **kwargs)
Return white top hat of an image.

The white top hat of an image is defined as the image minus its
morphological opening. This operation returns the bright spots of the image
that are smaller than the structuring element.

Parameters
----------
image : ndarray
Image array.
selem : ndarray, optional
The neighborhood expressed as an array of 1's and 0's.
If None, use cross-shaped structuring element (connectivity=1).
out : ndarray, optional
The array to store the result of the morphology. If None
is passed, a new array will be allocated.

Returns
-------
out : array, same shape and type as `image`
The result of the morphological white top hat.

Examples
--------
>>> # Subtract grey background from bright peak
>>> import numpy as np
>>> from skimage.morphology import square
>>> bright_on_grey = np.array([[2, 3, 3, 3, 2],
...                            [3, 4, 5, 4, 3],
...                            [3, 5, 9, 5, 3],
...                            [3, 4, 5, 4, 3],
...                            [2, 3, 3, 3, 2]], dtype=np.uint8)
>>> white_tophat(bright_on_grey, square(3))
array([[0, 0, 0, 0, 0],
[0, 0, 1, 0, 0],
[0, 1, 5, 1, 0],
[0, 0, 1, 0, 0],
[0, 0, 0, 0, 0]], dtype=uint8)

DATA
__all__ = ['binary_erosion', 'binary_dilation', 'binary_opening', 'bin...

这个也真够长的，也很详细。

想看看所有函数有哪些？

dir(skimage.morphology)

['__all__',
'__builtins__',
'__doc__',
'__file__',
'__name__',
'__package__',
'__path__',
'_convex_hull',
'_skeletonize',
'_skeletonize_3d',
'_skeletonize_3d_cy',
'_skeletonize_cy',
'_watershed',
'ball',
'binary',
'binary_closing',
'binary_dilation',
'binary_erosion',
'binary_opening',
'black_tophat',
'closing',
'convex_hull',
'convex_hull_image',
'convex_hull_object',
'cube',
'diamond',
'dilation',
'disk',
'erosion',
'grey',
'greyreconstruct',
'label',
'medial_axis',
'misc',
'octagon',
'octahedron',
'opening',
'reconstruction',
'rectangle',
'remove_small_holes',
'remove_small_objects',
'selem',
'skeletonize',
'skeletonize_3d',
'square',
'star',
'watershed',
'white_tophat']