Read and write images

Two easy ways: PIL and matplotlib (which uses PIL behind the scenes anyway). For PIL use: conda install pillow. Pillow is the new PIL.

`matplotlib`



In [1]:

    
import matplotlib.image as mpim



In [2]:

    
img = mpim.imread('../data/fluid.png')

We end up with a 4-channel array with (R,G,B) values in the range [0, 1]:



In [3]:

    
img









    Out[3]:





array([[[ 0.        ,  0.        ,  0.57254905,  1.        ],
        [ 0.        ,  0.        ,  0.57254905,  1.        ],
        [ 0.        ,  0.        ,  0.57254905,  1.        ],
        ..., 
        [ 0.        ,  0.        ,  0.57254905,  1.        ],
        [ 0.        ,  0.        ,  0.57254905,  1.        ],
        [ 0.        ,  0.        ,  0.57254905,  1.        ]],

       [[ 0.        ,  0.        ,  0.57254905,  1.        ],
        [ 0.        ,  0.        ,  0.57254905,  1.        ],
        [ 0.        ,  0.        ,  0.57254905,  1.        ],
        ..., 
        [ 0.        ,  0.        ,  0.57254905,  1.        ],
        [ 0.        ,  0.        ,  0.57254905,  1.        ],
        [ 0.        ,  0.        ,  0.57254905,  1.        ]],

       [[ 0.        ,  0.        ,  0.57254905,  1.        ],
        [ 0.        ,  0.        ,  0.57254905,  1.        ],
        [ 0.        ,  0.        ,  0.57254905,  1.        ],
        ..., 
        [ 0.        ,  0.        ,  0.57254905,  1.        ],
        [ 0.        ,  0.        ,  0.57254905,  1.        ],
        [ 0.        ,  0.        ,  0.57254905,  1.        ]],

       ..., 
       [[ 0.        ,  0.        ,  0.57254905,  1.        ],
        [ 0.        ,  0.        ,  0.57254905,  1.        ],
        [ 0.        ,  0.        ,  0.57254905,  1.        ],
        ..., 
        [ 0.        ,  0.        ,  0.57254905,  1.        ],
        [ 0.        ,  0.        ,  0.57254905,  1.        ],
        [ 0.        ,  0.        ,  0.57254905,  1.        ]],

       [[ 0.        ,  0.        ,  0.57254905,  1.        ],
        [ 0.        ,  0.        ,  0.57254905,  1.        ],
        [ 0.        ,  0.        ,  0.57254905,  1.        ],
        ..., 
        [ 0.        ,  0.        ,  0.57254905,  1.        ],
        [ 0.        ,  0.        ,  0.57254905,  1.        ],
        [ 0.        ,  0.        ,  0.57254905,  1.        ]],

       [[ 0.        ,  0.        ,  0.57254905,  1.        ],
        [ 0.        ,  0.        ,  0.57254905,  1.        ],
        [ 0.        ,  0.        ,  0.57254905,  1.        ],
        ..., 
        [ 0.        ,  0.        ,  0.57254905,  1.        ],
        [ 0.        ,  0.        ,  0.57254905,  1.        ],
        [ 0.        ,  0.        ,  0.57254905,  1.        ]]], dtype=float32)

plt.imshow() interprets the multiple channels as (R,G,B) values, so don't bother passing a colourmap:



In [4]:

    
import matplotlib.pyplot as plt
%matplotlib inline

plt.imshow(img, cmap='gray')
plt.show()

`pillow`

Even though you installed pillow, you use the PIL library. Weird, I know.



In [5]:

    
from PIL import Image



In [6]:

    
img = Image.open('../data/fluid.png')

The repr() of a PIL Image object renders the actual image in the notebook:



In [7]:

    
img









    Out[7]:



In [8]:

    
import numpy as np

i = np.array(img)

This time the array is represented as uint8 values, so [0,255].



In [9]:

    
i









    Out[9]:





array([[[  0,   0, 146, 255],
        [  0,   0, 146, 255],
        [  0,   0, 146, 255],
        ..., 
        [  0,   0, 146, 255],
        [  0,   0, 146, 255],
        [  0,   0, 146, 255]],

       [[  0,   0, 146, 255],
        [  0,   0, 146, 255],
        [  0,   0, 146, 255],
        ..., 
        [  0,   0, 146, 255],
        [  0,   0, 146, 255],
        [  0,   0, 146, 255]],

       [[  0,   0, 146, 255],
        [  0,   0, 146, 255],
        [  0,   0, 146, 255],
        ..., 
        [  0,   0, 146, 255],
        [  0,   0, 146, 255],
        [  0,   0, 146, 255]],

       ..., 
       [[  0,   0, 146, 255],
        [  0,   0, 146, 255],
        [  0,   0, 146, 255],
        ..., 
        [  0,   0, 146, 255],
        [  0,   0, 146, 255],
        [  0,   0, 146, 255]],

       [[  0,   0, 146, 255],
        [  0,   0, 146, 255],
        [  0,   0, 146, 255],
        ..., 
        [  0,   0, 146, 255],
        [  0,   0, 146, 255],
        [  0,   0, 146, 255]],

       [[  0,   0, 146, 255],
        [  0,   0, 146, 255],
        [  0,   0, 146, 255],
        ..., 
        [  0,   0, 146, 255],
        [  0,   0, 146, 255],
        [  0,   0, 146, 255]]], dtype=uint8)



In [ ]: