CODEWITHSUNDEEP
pythonarraysnumpy
safa
safa

Reputation: 65

Numpy: add row and column

How can I add one row and one column to a numpy array. The array has the shape (480,639,3) and I want to have the shape (481,640,3). The new row and column should filled with zeros, like this:

[43,42,40],         ...        [64,63,61], [0,0,0]
   ...              ...            ...     [0,0,0]
[29,29,29],         ...        [38,37,35], [0,0,0]
[0,0,0], [0,0,0]    ...                    [0,0,0]

To add a new column I'm doing this:

b = numpy.zeros((480,640,3), dtype = int)
b[:,:-1] = old_arry

But how I can add one row? Have I to use a loop or exists a better way to do this?

Upvotes: 2

Views: 1293

Answers (3)

hpaulj
hpaulj

Reputation: 231665

You can expand both dimensions at once, or you can expand it one dimension at a time. The operation is the same - create the target array, and copy the old to an appropriate slice. pad does this under the hood.

b = numpy.zeros((481,640,3), dtype = int)
b[:-1,:-1,:] = old_arry

for example:

In [527]: x=np.ones((2,2,3))
In [528]: y=np.zeros((3,3,3))
In [529]: y[:-1,:-1:,:]=x
In [530]: y
Out[530]: 
array([[[ 1.,  1.,  1.],
        [ 1.,  1.,  1.],
        [ 0.,  0.,  0.]],

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

       [[ 0.,  0.,  0.],
        [ 0.,  0.,  0.],
        [ 0.,  0.,  0.]]])

Upvotes: 0

Andrey Sobolev
Andrey Sobolev

Reputation: 12713

You can use np.vstack to stack arrays row-wise and np.hstack to stack them column-wise. See the code:

>>> import numpy as np
>>> x = np.arange(480*639*3).reshape((480,639,3))
>>> new_row = np.zeros((639,3))
>>> x = np.vstack((x,new_row[np.newaxis,:,:]))
>>> x.shape
(481, 639, 3)
>>> new_col = np.zeros((481,3))
>>> x = np.hstack([x, new_col[:,np.newaxis,:]])
>>> x.shape
(481, 640, 3)

Upvotes: 1

YXD
YXD

Reputation: 32521

You can use pad

>>> old = np.random.random_integers(0, 100, size=(480, 640))
>>> np.pad(old, pad_width=((0, 1), (0, 1)), mode='constant')
array([[ 66,  22,  51, ...,  18,  15,   0],
       [ 28,  12,  43, ...,   8,  38,   0],
       [ 55,  43,  89, ...,  67,  58,   0],
       ...,
       [ 17,  25, 100, ...,  12,  52,   0],
       [ 97,  59,  82, ...,  38,  97,   0],
       [  0,   0,   0, ...,   0,   0,   0]])
>>> np.pad(old, pad_width=((0, 1), (0, 1)), mode='constant').shape
(481, 641)
>>>

You can also write it as np.pad(old, ((0, 1), (0, 1)), mode='constant'), i.e. without the pad_width keyword. To set a different value for the padded areas, see the constant_values parameter in the documentation.

Upvotes: 3

Related Questions