使用模式P将阵列转换为1通道图像

Question

I'm trying to convert a numpy array with (0,10) values to a 1-channel colored image

Example:

result = [[0 0 1]
          [0 3 1]
          [1 2 2]]

to:

I tried to use this code:

cm = ListedColormap(color_map(4, True), 'pascal', 4)
plt.register_cmap(cmap=cm)
plt.imsave('outputarray.png', result, cmap='pascal')

(color_map is from: https://gist.github.com/wllhf/a4533e0adebe57e3ed06d4b50c8419ae )

But then:

im = Image.open('outputarray.png')
im2 = np.array(im)
print im2.shape
print im2.min()
print im2.max()

returns:

shape: (3, 3, 4)
min: 0
max: 255

I supposed it should be:

shape: (3, 3)
min: 0
max: 3

Thanks!

Answer 1

The shape is (3, 3, 4) since the image is a png image which has four channels, rgb and alpha.

In order to obtain the original array shape, you can convert the image to grayscale which only has one channel.

>>> im2 = im.convert("L")
>>> im2 = np.array(im2)
>>> im2
array([[  0,   0,  38],
       [  0, 113,  38],
       [ 38,  75,  75]], dtype=uint8)
>>> im2.shape
(3, 3)

And then replace the grayscale values with the smallest possible integer.

>>> for i in range(3, 0, -1):
...     im2[im2==np.max(im2)] = i
>>> im2
array([[0, 0, 1],
       [0, 3, 1],
       [1, 2, 2]], dtype=uint8)
>>> im2.min
0
>>> im2.max
3