5 Best Ways to Convert a Python NumPy Array to JPEG

by
💡 Problem Formulation: In various domains such as data analysis, scientific computing, or machine learning, it’s common to handle image data as NumPy arrays in Python. However, you may need to store or display these arrays as standard JPEG images. This article guides you through five different methods to convert a NumPy array that represents image data into a JPEG file. For example, we’ll start with an input of a NumPy array with pixel values and aim to create a JPEG image as the desired output.

Method 1: Using PIL/Pillow

Python Imaging Library (PIL) and its fork, Pillow, provide extensive file format support, efficient internal representation, and fairly powerful image processing capabilities. To convert a NumPy array to a JPEG image, simply use the Image.fromarray() method from PIL/Pillow, which creates an image memory from an object exporting the array interface, such as a NumPy array. Then, you can use the save() function specifying ‘JPEG’ as the format.

Here’s an example:

♥️ Info: Are you AI curious but you still have to create real impactful projects? Join our official AI builder club on Skool (only $5): SHIP! - One Project Per Month

from PIL import Image
import numpy as np

array = np.zeros((100, 100, 3), dtype=np.uint8)
img = Image.fromarray(array, 'RGB')
img.save('output.jpeg')

Output: ‘output.jpeg’ – A JPEG image file with a 100×100 black square.

This snippet creates a 100×100 pixel black square by first generating a zero-initialized NumPy array with shape (100, 100, 3) for the RGB channels. It then converts this array into a PIL Image object and finally saves it as a JPEG file named ‘output.jpeg’.

Method 2: Using Matplotlib

Matplotlib is a popular plotting library for Python which also contains functions to save images. With its imsave() function, you get a convenient way to save a NumPy array as an image file. Specify the filename and array, and Matplotlib handles the conversion to a JPEG image internally.

Here’s an example:

import matplotlib.pyplot as plt
import numpy as np

array = np.random.rand(100, 100, 3)
plt.imsave('random_output.jpeg', array)

Output: ‘random_output.jpeg’ – A JPEG image file with randomly colored pixels.

The code generates a 100×100 pixel image filled with random colors by creating a NumPy array with random values between 0 and 1 in each channel. The plt.imsave() function then takes the array and saves it directly as a ‘random_output.jpeg’ file without the need to display it first.

Method 3: Using OpenCV

OpenCV is a robust framework for computer vision tasks, which includes features for image processing and file I/O. The cv2.imwrite() function is perfect for writing NumPy arrays as image files, including the JPEG format. Simply supply the filename with a ‘.jpeg’ extension and the array.

Here’s an example:

import cv2
import numpy as np

array = np.full((100, 100, 3), 255, dtype=np.uint8)
cv2.imwrite('white_output.jpeg', array)

Output: ‘white_output.jpeg’ – A JPEG image file showing a 100×100 white square.

This code creates a white 100×100 pixel image. A white color is represented by setting all color channels to maximum, which is done here by filling the NumPy array with 255s. The cv2.imwrite() function then writes the array to a ‘white_output.jpeg’ file.

Method 4: Using Imageio

Imageio is a Python library that provides an easy interface to read and write a wide range of image data, including animated images, volumetric data, and scientific formats. It’s as straightforward as calling the imageio.imwrite() function with a filename that ends with ‘.jpg’ or ‘.jpeg’ and the NumPy array.

Here’s an example:

import imageio
import numpy as np

array = np.random.randint(0, 256, (100, 100, 3), dtype=np.uint8)
imageio.imwrite('colorful_output.jpeg', array)

Output: ‘colorful_output.jpeg’ – A JPEG file with a 100×100 array of random colors.

The snippet creates an image filled with random pixel colors by using the np.random.randint() function to fill a NumPy array with random values in the range 0 to 255 for each color channel. The imageio.imwrite() method then saves this array as a ‘colorful_output.jpeg’.

Bonus One-Liner Method 5: Using SciPy

As part of its miscellaneous utility functions, SciPy used to offer scipy.misc.imsave(), which has since been deprecated, but for legacy purposes, it offers a very concise way to save a NumPy array as an image. This method is generally not recommended for new code.

Here’s an example:

# Note: 'scipy.misc.imsave()' is deprecated.
# Use 'imageio.imwrite()' as an alternative.
import scipy.misc
import numpy as np

array = np.ones((100, 100, 3), dtype=np.uint8) * 127
scipy.misc.imsave('gray_output.jpeg', array)

Output: ‘gray_output.jpeg’ – A JPEG file showing a 100×100 gray square.

Even though it’s shown for informative purposes, this example uses the scipy.misc.imsave() function to save a NumPy array with all values set to 127, resulting in a gray square, as a JPEG image.

Summary/Discussion

  • Method 1: PIL/Pillow. Great for simplicity and additional image processing. However, requires installation of a third-party library.
  • Method 2: Matplotlib. A quick option for those already using Matplotlib for plotting. Doesn’t always produce the smallest file sizes.
  • Method 3: OpenCV. Ideal for computer vision projects. May be an overkill for simple image-saving purposes.
  • Method 4: Imageio. Simple and powerful tool for image reading/writing tasks. Supports a wide array of formats.
  • Bonus Method 5: SciPy. Provided as a one-liner legacy method; however, not recommended for use in new projects due to its deprecation.