Read in data using scipy (image)

Read in data using scipy (image)

IMPORTANT: This walk through makes use of a few additional packages that work with numpy arrays, namely scipy and matplotlib

As seen here, numpy can be used with tabulated data sets but where it really excels is when it is applied to workflows involving matrices of data - think of photographs, satellite images, digital elevation models (DEMs) etc.

Right click this link and save this file to a location on your machine.

Open an image as a numpy array

To read the image in as a numpy array, we are going to need a little help from a function from the scipy library:

from scipy import misc

You now have access to the routines available within scipy’s misc module. To read in the image, type the following:


You have now created a numpy array object holding the image information. You can see the image in its array form using:


Plot your image

You can visualise the array using pyplot from the matplotlib module:

import matplotlib.pyplot

which will result in:

"RGB puppy"

By using shape you will see that the array has 3 dimensions - as the file we are using is an RGB image, in this case, each dimension applies to its red, green and blue colour bands:

>>> dog.shape
>>> (400L, 600L, 3L)

To slice the array and access the red component use:


for the green component use:


and for the blue component use:


Now that we have accessed slices of the image, let’s manipulate them and create something new. Lets turn our RGB image into a greyscale image. This can be achieved using the equation:

grey = (0.2126 * red) + (0.7152 * green) + (0.0722 * blue)

To make a greyscale array, we’ll aply the above equation and use the sliced parts of the original image as held by the arrays we called red, green and blue. We’ll call our new array grey:

grey = (0.2126 * red) + (0.7152 * green) + (0.0722 * blue)

No looping or anything required - this efficiency is what makes numpy fast. If we had used a vanilla python approach, we would have to have looped through each element of the array which would have taken much longer.

Now we have our grey image, we can plot it using matplotlib again, but this time we must tell it that the image is greyscale - this is done by using the cmap option within the imshow command:

import as cm 
matplotlib.pyplot.imshow(grey, cmap =

To write the array out as an image, we’ll use scipy again:

misc.imsave('puppy_grey.png', grey)

Open this image and check out your funky new colour scheme, all thanks to a bit of python and numpy.

Previous Home Next