4

我有一个二维数组，a包含一组 100 个 x,y,z 坐标：

[[ 0.81  0.23  0.52]
 [ 0.63  0.45  0.13]
 ...
 [ 0.51  0.41  0.65]]

我想创建一个 3D 二值图像，b在每个 x、y、z 维度上都有 101 个像素，坐标范围在 0.00 到 1.00 之间。由 定义的位置的a像素应取值为 1，所有其他像素的值应为 0。

我可以用 来创建一个正确形状的零数组b = np.zeros((101,101,101))，但是如何分配坐标并将其切片以使用a？

4

2 回答 2

4

你可以做这样的事情 -

# Get the XYZ indices
idx = np.round(100 * a).astype(int)

# Initialize o/p array
b = np.zeros((101,101,101))

# Assign into o/p array based on linear index equivalents from indices array
np.put(b,np.ravel_multi_index(idx.T,b.shape),1)

分配部分的运行时 -

让我们使用更大的网格来进行计时。

In [82]: # Setup input and get indices array
    ...: a = np.random.randint(0,401,(100000,3))/400.0
    ...: idx = np.round(400 * a).astype(int)
    ...: 

In [83]: b = np.zeros((401,401,401))

In [84]: %timeit b[list(idx.T)] = 1 #@Praveen soln
The slowest run took 42.16 times longer than the fastest. This could mean that an intermediate result is being cached.
1 loop, best of 3: 6.28 ms per loop

In [85]: b = np.zeros((401,401,401))

In [86]: %timeit np.put(b,np.ravel_multi_index(idx.T,b.shape),1) # From this post
The slowest run took 45.34 times longer than the fastest. This could mean that an intermediate result is being cached.
1 loop, best of 3: 5.71 ms per loop

In [87]: b = np.zeros((401,401,401))

In [88]: %timeit b[idx[:,0],idx[:,1],idx[:,2]] = 1 #Subscripted indexing
The slowest run took 40.48 times longer than the fastest. This could mean that an intermediate result is being cached.
1 loop, best of 3: 6.38 ms per loop
于 2016-08-30T05:58:42.350 回答
4

首先，从安全地将浮点数舍入为整数开始。在上下文中，请参阅此问题。

a_indices = np.rint(a * 100).astype(int)

接下来，将这些索引分配b给 1。但要小心使用普通list数组而不是数组，否则会触发索引数组的使用。似乎这种方法的性能与替代方法的性能相当（感谢@Divakar！:-)

b[list(a_indices.T)] = 1

我创建了一个大小为 10 而不是 100 和 2 维而不是 3 的小示例来说明：

>>> a = np.array([[0.8, 0.2], [0.6, 0.4], [0.5, 0.6]])
>>> a_indices = np.rint(a * 10).astype(int)
>>> b = np.zeros((10, 10))
>>> b[list(a_indices.T)] = 1
>>> print(b) 
[[ 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.  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.  1.  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.  0.  0.  0.  0.  0.  0.  0.  0.  0.]]
于 2016-08-30T06:07:10.253 回答