Python⾃带slic代码分析⼀.python中的slic函数
def slic(image, n_gments=100, compactness=10., max_iter=10, sigma=0,
spacing=None, multichannel=True, convert2lab=None,
e blocks
enforce_connectivity=True, min_size_factor=0.5, max_size_factor=3,
slic_zero=Fal):
"""Segments image using k-means clustering in Color-(x,y,z) space.
Parameters
----------
image : 2D, 3D or 4D ndarray
Input image, which can be 2D or 3D, and grayscale or multichannel
(e `multichannel` parameter).
n_gments : int, optional
The (approximate) number of labels in the gmented output image.
compactness : float, optional
控制颜⾊和空间之间的平衡，约⾼越⽅块，和图关系密切，最好先确定指数级别，再微调
Balances color proximity and space proximity. Higher values give
more weight to space proximity, making superpixel shapes more
square/cubic. In SLICO mode, this is the initial compactness.
This parameter depends strongly on image contrast and on the
shapes of objects in the image. We recommend exploring possible
values on a log scale, e.g., 0.01, 0.1, 1, 10, 100, before
refining around a chon value.
aige
max_iter : int, optional
最⼤k均值迭代次数
Maximum number of iterations of k-means.
sigma : float or (3,) array-like of floats, optional
图像每个维度进⾏预处理时的⾼斯平滑核宽。若给定为标量值，则同⼀个值运⽤到各个维度。0意味
着不平滑。如果“sigma”是标量的，并且提供了⼿动体素间距，则⾃动缩放它（参见注释部分）。
Width of Gaussian smoothing kernel for pre-processing for each
dimension of the image. The same sigma is applied to each dimension in
ca of a scalar value. Zero means no smoothing.
Note, that `sigma` is automatically scaled if it is scalar and a
manual voxel spacing is provided (e Notes ction).
spacing : (3,) array-like of floats, optional
代表沿着图像每个维度的体素空间。默认情况下，slic假定均匀的空间（沿x,y,z轴相同的体素分辨
率），这个参数控制在k均值聚类中各轴距离的权重
The voxel spacing along each image dimension. By default, `slic`
assumes uniform spacing (same voxel resolution along z, y and x).
This parameter controls the weights of the distances along z, y,
san diego
and x during k-means clustering.
multichannel : bool, optional
⼆进制参数，代表图像的最后⼀个轴代表多通道还是另⼀个空间维度
Whether the last axis of the image is to be interpreted as multiple
channels or another spatial dimension.
convert2lab : bool, optional
⼆进制参数，判断输⼊需要在分割之前转到LAB颜⾊空间。输⼊必须是RGB。当多通道参数为True，输⼊图⽚的通道数为3时，该参数默认为True
Whether the input should be converted to Lab colorspace prior to
gmentation. The input image *must* be RGB. Highly recommended.
This option defaults to ``True`` when ``multichannel=True`` *and*
``image.shape[-1] == 3``.
enforce_connectivity: bool, optional
⼆进制参数，控制⽣成的分割块连接或不连接
Whether the generated gments are connected or not
min_size_factor: float, optional
与分割⽬标数有关的要删去的最⼩分割块⽐率，（⼤概是⼩于长*宽*⾼/⽬标数量的分割结果会被融
合掉）
Proportion of the minimum gment size to be removed with respect
to the suppod gment size ```depth*width*height/n_gments```
max_size_factor: float, optional
max_size_factor: float, optional
最⼤融合⽐率上限
Proportion of the maximum connected gment size. A value of 3 works
in most of the cas.
slic_zero: bool, optional
不知所谓的零参数
Run SLIC-zero, the zero-parameter mode of SLIC. [2]_
Returns
-------
labels : 2D or 3D array
Integer mask indicating gment labels.
suck是什么意思
Rais
------
ValueError
If ``convert2lab`` is t to ``True`` but the last array
dimension is not of length 3.
Notes
-----
who care* If `sigma > 0`, the image is smoothed using a Gaussian kernel prior to
gmentation.
* If `sigma` is scalar and `spacing` is provided, the kernel width is
divided along each dimension by the spacing. For example, if ``sigma=1``
and ``spacing=[5, 1, 1]``, the effective `sigma` is ``[0.2, 1, 1]``. This
音标翻译
ensures nsible smoothing for anisotropic images.
如果有平滑参数sigma和体素空间参数spacing，那么空间体素参数会对平滑参数有平分的影响，⽐如 1/[5,1,1]=[0.2,1,1]
* The image is rescaled to be in [0, 1] prior to processing.
图像在预处理之前会被处理为[0,1]之间的标量
* Images of shape (M, N, 3) are interpreted as 2D RGB images by default. To
interpret them as 3D with the last dimension having length 3, u
`multichannel=Fal`.
（M,N,3）的图像默认为2维（RGB的图像），要想被理解为3维图需要设置多通道参数=Fal
References
----------
.. [1] Radhakrishna Achanta, Appu Shaji, Kevin Smith, Aurelien Lucchi,
Pascal Fua, and Sabine Süsstrunk, SLIC Superpixels Compared to
State-of-the-art Superpixel Methods, TPAMI, May 2012.
平面设计培训学校.. [2] ivrg.epfl.ch/rearch/superpixels#SLICO
Examples
--------
>>> ation import slic
>>> from skimage.data import astronaut
>>> img = astronaut()
>>> gments = slic(img, n_gments=100, compactness=10)
Increasing the compactness parameter yields more square regions:
>>> gments = slic(img, n_gments=100, compactness=20)
"""
>>>>>>>>>##⼲正事啦image = img_as_float(image)
is_2d = Fal
#2D灰度图
if image.ndim == 2:
# 2D grayscale image
image = waxis, ..., np.newaxis]
is_2d = True
#⽐如2D RGB的图
elif image.ndim == 3 and multichannel:
# Make 2D multichannel image 3D with depth = 1
image = waxis, ...]
is_2d = True
#⽐如3D图
elif image.ndim == 3 and not multichannel:
elif image.ndim == 3 and not multichannel:
# Add channel as single last dimension
image = image[..., np.newaxis]
#控制聚类时各轴权重
if spacing is None:
spacing = np.ones(3)
elif isinstance(spacing, (list, tuple)):
spacing = np.array(spacing, dtype=np.double)
#⾼斯平滑
employee是什么意思if not isinstance(sigma, coll.Iterable):
sigma = np.array([sigma, sigma, sigma], dtype=np.double)
sigma /= spacing.astype(np.double)#有可能发⽣的体素除
elif isinstance(sigma, (list, tuple)):
sigma = np.array(sigma, dtype=np.double)
#⾼斯滤波处
if (sigma > 0).any():
# add zero smoothing for multichannel dimension
sigma = list(sigma) + [0]
image = ndi.gaussian_filter(image, sigma)
#多通道RGB图且需要转lab，⽤rab2lab即可实现
if multichannel and (convert2lab or convert2lab is None):
if image.shape[-1] != 3 and convert2lab:
rai ValueError("Lab colorspace conversion requires a RGB image.") elif image.shape[-1] == 3:
image = rgb2lab(image)
depth, height, width = image.shape[:3]
# initialize cluster centroids for desired number of gments
#为实现⽬标分割块数，初始化聚类中⼼。喜羊羊与灰太狼之给快乐加
#grid_* 相当于index
#slices是根据⽬标数量分的块，有取整需要
grid_z, grid_y, grid_x = np.mgrid[:depth, :height, :width]
slices = regular_grid(image.shape[:3], n_gments)
step_z, step_y, step_x = [int(s.step if s.step is not None el 1)
for s in slices]
gments_z = grid_z[slices]
gments_y = grid_y[slices]
gments_x = grid_x[slices]
gments_color = np.zeros(gments_z.shape + (image.shape[3],))
gments = np.concatenate([gments_z[..., np.newaxis],
gments_y[..., np.newaxis],
gments_x[..., np.newaxis],
gments_color],
axis=-1).reshape(-1, 3 + image.shape[3])
gments = np.ascontiguousarray(gments)
# we do the scaling of ratio in the same way as in the SLIC paper
# so the values have the same meaning
step = float(max((step_z, step_y, step_x)))
ratio = 1.0 / compactness
#我类个去，分割时⽅不⽅的骚操作
image = np.ascontiguousarray(image * ratio)
labels = _slic_cython(image, gments, step, max_iter, spacing, slic_zero)
#把过⼩过⼩的处理⼀下
if enforce_connectivity:
gment_size = depth * height * width / n_gments
min_size = int(min_size_factor * gment_size)
min_size = int(min_size_factor * gment_size)
max_size = int(max_size_factor * gment_size)
labels = _enforce_label_connectivity_cython(labels,
min_size,
max_size)
if is_2d:
labels = labels[0]
vaguelyreturn labels
⼆、注意事项
1.要不要转化到LAB空间是可以调的，当然啦不转的结果就是⽅⽅正正的空间分割，和内容毫⽆关系，⽐如下图
convert2lab or convert2lab is None 这段代码可以看出来，不传参数和传参数为True都是转到lab了，完美。
2.分割结果⽐设置的少是因为做了⼀下后处理，调⼀下enforce_connectivity就变多啦，但不是⼀定分割结果和设置数量⼀样的。⽐如下图，设置分割20，参数设置为Fal结果为12块，参数设置为True就是3块
分割⽬标数量恰当的话（⽐如100），是这样的：
