Saving a Numpy Array as an Image Without PIL

Introduction

In data processing and computer vision, it is often necessary to visualize numerical data by converting it into images. This can be particularly useful for visualizing datasets or debugging algorithms that process image data. In this tutorial, we will explore how to save a Numpy array as an image file without using the Python Imaging Library (PIL). We’ll cover several methods including using Matplotlib, OpenCV, and even writing PNG files manually in pure Python.

Prerequisites

Before proceeding with this tutorial, ensure you have:

A basic understanding of Python programming.
The numpy library installed (pip install numpy).
Familiarity with image formats like PNG and JPEG.

We will also use some additional libraries, which can be installed as follows:

pip install matplotlib opencv-python-headless

Method 1: Using Matplotlib

Matplotlib is a powerful plotting library that also supports saving images directly from Numpy arrays. This method is straightforward and effective for simple image saving tasks.

Steps to Save an Image with Matplotlib

Import Required Libraries:

import numpy as np
import matplotlib.pyplot as plt

Create or Load a Numpy Array:

Suppose you have a 256×256 grayscale image stored in a Numpy array.

# Example: Create a sample image using a gradient
data = np.linspace(0, 1, 256).reshape((256, -1))

Save the Array as an Image:

Use plt.imsave() to save the array in your desired format (e.g., PNG).
```
plt.imsave('output_image.png', data)
```

Method 2: Using OpenCV

OpenCV is a robust library for computer vision tasks and provides efficient ways to manipulate images, including saving them from Numpy arrays.

Steps to Save an Image with OpenCV

Import Required Libraries:
```
import cv2
import numpy as np
```
Create or Load a Numpy Array:

You can use any existing array or generate one, similar to the Matplotlib example.
Save the Image Using OpenCV:

Use cv2.imwrite() to save the image file.
```
# Assuming 'data' is your image in a numpy array format
cv2.imwrite('output_image.png', (data * 255).astype(np.uint8))
```
Note: OpenCV uses BGR by default. If working with RGB data, convert it using cv2.cvtColor().

Method 3: Writing PNG Manually in Pure Python

For environments where you can’t rely on external libraries like Matplotlib or OpenCV, you can write a custom function to save images as PNG files directly from Numpy arrays. This method provides flexibility and control over the image saving process.

Steps to Write a PNG File

Define the Function:

Here’s a function that writes a compressed RGBA PNG file using only standard Python libraries:

def write_png(buf, width, height):
    import zlib, struct
    
    # Reverse 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)))

    # Prepare PNG file structure
    return b''.join([
        b'\x89PNG\r\n\x1a\n',
        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'')])

Save the Image:

Create a buffer from your Numpy array and use the function to write it as a PNG file.

# Example: Convert a grayscale image (256x256) to RGBA
data = np.random.rand(256, 256)
buf = (data * 255).astype(np.uint8).tobytes()

png_data = write_png(buf, 256, 256)

with open("manual_image.png", 'wb') as fh:
    fh.write(png_data)

Conclusion

This tutorial provided multiple approaches to save Numpy arrays as image files without relying on PIL. Whether you choose Matplotlib for its simplicity and visualization capabilities, OpenCV for more complex processing tasks, or pure Python for flexibility in constrained environments, each method has its own advantages. By understanding these techniques, you can effectively manage image data within your projects.