Home » Python » Saving a Numpy array as an image

Saving a Numpy array as an image

Posted by: admin November 1, 2017 Leave a comment


I have a matrix in the type of a Numpy array. How would I write it to disk it as an image? Any format works (png, jpeg, bmp…). One important constraint is that PIL is not present.


You can use PyPNG. It’s a pure Python (no dependencies) open source PNG encoder/decoder and it supports writing NumPy arrays as images.


This uses PIL, but maybe some might find it useful:

import scipy.misc
scipy.misc.imsave('outfile.jpg', image_array)

EDIT: The current scipy version started to normalize all images so that min(data) become black and max(data) become white. This is unwanted if the data should be exact grey levels or exact RGB channels. The solution:

import scipy.misc
scipy.misc.toimage(image_array, cmin=0.0, cmax=...).save('outfile.jpg')


An answer using PIL (just in case it’s useful).

given a numpy array “A”:

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

you can replace “jpeg” with almost any format you want. More details about the formats here


Pure Python (2 & 3), a snippet without 3rd party dependencies.

This function writes compressed, true-color (4 bytes per pixel) RGBA PNG’s.

def write_png(buf, width, height):
    """ buf: must be bytes or a bytearray in Python3.x,
        a regular string in Python2.x.
    import zlib, struct

    # reverse the vertical line order and add null bytes at the start
    width_byte_4 = width * 4
    raw_data = b''.join(b'\x00' + buf[span:span + width_byte_4]
                        for span in range((height - 1) * width_byte_4, -1, - width_byte_4))

    def png_pack(png_tag, data):
        chunk_head = png_tag + data
        return (struct.pack("!I", len(data)) +
                chunk_head +
                struct.pack("!I", 0xFFFFFFFF & zlib.crc32(chunk_head)))

    return b''.join([
        png_pack(b'IHDR', struct.pack("!2I5B", width, height, 8, 6, 0, 0, 0)),
        png_pack(b'IDAT', zlib.compress(raw_data, 9)),
        png_pack(b'IEND', b'')])

… The data should be written directly to a file opened as binary, as in:

data = write_png(buf, 64, 64)
with open("my_image.png", 'wb') as fd:


With matplotlib:

import matplotlib

matplotlib.image.imsave('name.png', array)

Works with matplotlib 1.3.1, I don’t know about lower version. From the docstring:

    A string containing a path to a filename, or a Python file-like object.
    If *format* is *None* and *fname* is a string, the output
    format is deduced from the extension of the filename.
    An MxN (luminance), MxNx3 (RGB) or MxNx4 (RGBA) array.

enter image description here


If you have matplotlib, you can do:

import matplotlib.pyplot as plt
plt.imshow(matrix) #Needs to be in row,col order

This will save the plot (not the images itself).
enter image description here


There’s opencv for python (http://docs.opencv.org/trunk/doc/py_tutorials/py_tutorials.html).

import cv2
import numpy as np

cv2.imwrite("filename.png", np.zeros((10,10)))

useful if you need to do more processing other than saving.


Addendum to @ideasman42’s answer:

def saveAsPNG(array, filename):
    import struct
    if any([len(row) != len(array[0]) for row in array]):
        raise ValueError, "Array should have elements of equal size"

                                #First row becomes top row of image.
    flat = []; map(flat.extend, reversed(array))
                                 #Big-endian, unsigned 32-byte integer.
    buf = b''.join([struct.pack('>I', ((0xffFFff & i32)<<8)|(i32>>24) )
                    for i32 in flat])   #Rotate from ARGB to RGBA.

    data = write_png(buf, len(array[0]), len(array))
    f = open(filename, 'wb')

So you can do:

saveAsPNG([[0xffFF0000, 0xffFFFF00],
           [0xff00aa77, 0xff333333]], 'test_grid.png')

Producing test_grid.png:

Grid of red, yellow, dark-aqua, grey

(Transparency also works, by reducing the high byte from 0xff.)


matplotlib svn has a new function to save images as just an image — no axes etc. it’s a very simple function to backport too, if you don’t want to install svn (copied straight from image.py in matplotlib svn, removed the docstring for brevity):

def imsave(fname, arr, vmin=None, vmax=None, cmap=None, format=None, origin=None):
    from matplotlib.backends.backend_agg import FigureCanvasAgg as FigureCanvas
    from matplotlib.figure import Figure

    fig = Figure(figsize=arr.shape[::-1], dpi=1, frameon=False)
    canvas = FigureCanvas(fig)
    fig.figimage(arr, cmap=cmap, vmin=vmin, vmax=vmax, origin=origin)
    fig.savefig(fname, dpi=1, format=format)


The world probably doesn’t need yet another package for writing a numpy array to a PNG file, but for those who can’t get enough, I recently put up numpngw on github:


and on pypi: https://pypi.python.org/pypi/numpngw/

The only external dependency is numpy.

Here’s the first example from the examples directory of the repository. The essential line is simply

write_png('example1.png', img)

where img is a numpy array. All the code before that line is import statements and code to create img.

import numpy as np
from numpngw import write_png

# Example 1
# Create an 8-bit RGB image.

img = np.zeros((80, 128, 3), dtype=np.uint8)

grad = np.linspace(0, 255, img.shape[1])

img[:16, :, :] = 127
img[16:32, :, 0] = grad
img[32:48, :, 1] = grad[::-1]
img[48:64, :, 2] = grad
img[64:, :, :] = 127

write_png('example1.png', img)

Here’s the PNG file that it creates:



If you happen to use [Py]Qt already, you may be interested in qimage2ndarray. Starting with version 1.4 (just released), PySide is supported as well, and there will be a tiny imsave(filename, array) function similar to scipy’s, but using Qt instead of PIL. With 1.3, just use something like the following:

qImage = array2qimage(image, normalize = False) # create QImage from ndarray
success = qImage.save(filename) # use Qt's image IO functions for saving PNG/JPG/..

(Another advantage of 1.4 is that it is a pure python solution, which makes this even more lightweight.)


Assuming you want a grayscale image:

im = Image.new('L', (width, height))