Question

I have a NumPy array of 3,076,568 binary values (1s and 0s). I would like to convert this to a matrix, and then to a grayscale image in Python.

However, when I try to reshape the array into a 1,538,284 x 1,538,284 matrix, I get a memory error.

How can I reduce the size of the matrix so that it will turn into an image that will fit on a screen without losing the uniqueness/data?

Furthermore, how would I turn it into a grayscale image?

Any help or advice would be appreciated. Thank you.

Answer 1

Your array of “binary values” is an array of bytes?

If so, you can do (using Pillow) after resizing it:

from PIL import Image
im = Image.fromarray(arr)

And then im.show() to see it.

If your array has only 0’s and 1’s (1-bit depth or b/w) you may have to multiply it to 255

im = Image.fromarray(arr * 255)

Here an example:

>>> arr = numpy.random.randint(0,256, 100*100) #example of a 1-D array
>>> arr.resize((100,100))
>>> im = Image.fromarray(arr)
>>> im.show()

Edit (2018):

This question was written in 2011 and Pillow changed ever since requiring to use the mode='L' parameter when loading with fromarray.

Also on comments below it was said arr.astype(np.uint8) was needed as well, but I have not tested it