scipy distance_transform_edt函数如何工作？

Question

https://docs.scipy.org/doc/scipy-0.14.0/reference/generated/scipy.ndimage.morphology.distance_transform_edt.html

我无法理解欧几里德距离变换函数在Scipy中是如何工作的.据我所知,它与Matlab函数(bwdist)不同.作为一个例子,输入:

[[ 0.  0.  0.  0.  0.]
 [ 0.  1.  0.  0.  0.]
 [ 0.  0.  0.  0.  0.]
 [ 0.  0.  0.  1.  0.]
 [ 0.  0.  0.  0.  0.]]

scipy.ndimage.distance_transform_edt函数返回相同的数组:

[[ 0.  0.  0.  0.  0.]
 [ 0.  1.  0.  0.  0.]
 [ 0.  0.  0.  0.  0.]
 [ 0.  0.  0.  1.  0.]
 [ 0.  0.  0.  0.  0.]]

但matlab函数返回:

1.4142    1.0000    1.4142    2.2361    3.1623
1.0000         0    1.0000    2.0000    2.2361
1.4142    1.0000    1.4142    1.0000    1.4142
2.2361    2.0000    1.0000         0    1.0000
3.1623    2.2361    1.4142    1.0000    1.4142

这更有意义,因为它将"距离"返回到最近的那个.

Answer 1

从文档字符串中不清楚,但distance_transform_edt计算从非零(即非背景)点到最近的零(即背景)点的距离.

例如:

In [42]: x
Out[42]: 
array([[0, 0, 0, 0, 0, 1, 1, 1],
       [0, 1, 1, 1, 0, 1, 1, 1],
       [0, 1, 1, 1, 0, 1, 1, 1],
       [0, 0, 1, 1, 0, 0, 0, 1]])

In [43]: np.set_printoptions(precision=3)  # Easier to read the result with fewer digits.

In [44]: distance_transform_edt(x)
Out[44]: 
array([[ 0.   ,  0.   ,  0.   ,  0.   ,  0.   ,  1.   ,  2.   ,  3.   ],
       [ 0.   ,  1.   ,  1.   ,  1.   ,  0.   ,  1.   ,  2.   ,  2.236],
       [ 0.   ,  1.   ,  1.414,  1.   ,  0.   ,  1.   ,  1.   ,  1.414],
       [ 0.   ,  0.   ,  1.   ,  1.   ,  0.   ,  0.   ,  0.   ,  1.   ]])

你可以得到Matlab的相当于bwdist(a)通过应用distance_transform_edt()来np.logical_not(a)(即反转前景色和背景):

In [71]: a
Out[71]: 
array([[ 0.,  0.,  0.,  0.,  0.],
       [ 0.,  1.,  0.,  0.,  0.],
       [ 0.,  0.,  0.,  0.,  0.],
       [ 0.,  0.,  0.,  1.,  0.],
       [ 0.,  0.,  0.,  0.,  0.]])

In [72]: distance_transform_edt(np.logical_not(a))
Out[72]: 
array([[ 1.414,  1.   ,  1.414,  2.236,  3.162],
       [ 1.   ,  0.   ,  1.   ,  2.   ,  2.236],
       [ 1.414,  1.   ,  1.414,  1.   ,  1.414],
       [ 2.236,  2.   ,  1.   ,  0.   ,  1.   ],
       [ 3.162,  2.236,  1.414,  1.   ,  1.414]])