如何在Python中保存1 bit深度的PNG二进制图像？

Hey there! I totally get the frustration of trying to save a 1-bit PNG and hitting walls with PIL/Pillow or OpenCV. Let's break down exactly how to make this work for both libraries—you probably just missed a key step or parameter.

Using Pillow (PIL) to Save 1-bit PNGs #

This is the most straightforward method once you know the right conversion step. The key is ensuring your image is explicitly converted to the 1-bit mode before saving.

Step-by-Step Code: #

from PIL import Image

# Let's assume you already have your binary image—either loaded via Pillow or converted from a numpy array
# Example if starting from a numpy array:
# import numpy as np
# binary_np = np.array([[0, 255], [255, 0]], dtype=np.uint8)  # Sample binary image
# img = Image.fromarray(binary_np)

# Convert the image to 1-bit mode (critical step!)
# Use dither=Image.NONE if you want no dithering (strict black/white, no intermediate shades)
img_1bit = img.convert('1', dither=Image.NONE)

# Save as PNG—Pillow will handle the 1-bit encoding automatically
img_1bit.save('output_1bit_pillow.png', format='PNG')

Why this works: #

• The convert('1') method transforms your image into a 1-bit bilevel (black/white) mode. Without this, even if your image looks binary, it's likely stored as 8-bit grayscale ('L' mode), which is what you were probably saving before.
• Checking the mode afterward: Load the saved image back with Image.open('output_1bit_pillow.png') and print img.mode—it should return '1'.

Using OpenCV (With a Workaround) #

OpenCV doesn't natively support saving 1-bit PNGs directly, but we can either convert to a Pillow Image first (using the method above) or use the pypng library to write the 1-bit data directly.

Option 1: Convert to Pillow Image #

import cv2
from PIL import Image

# Load your binary image with OpenCV (grayscale mode)
cv_img = cv2.imread('input_binary.png', cv2.IMREAD_GRAYSCALE)

# Convert OpenCV image (grayscale) to Pillow Image
pillow_img = Image.fromarray(cv_img)

# Follow the Pillow steps above
img_1bit = pillow_img.convert('1', dither=Image.NONE)
img_1bit.save('output_1bit_cv2_pillow.png', format='PNG')

Option 2: Use the pypng Library #

If you prefer to stick with OpenCV without Pillow, install pypng first with pip install pypng, then use this code:

import cv2
import png

# Load binary grayscale image
cv_img = cv2.imread('input_binary.png', cv2.IMREAD_GRAYSCALE)

# Convert 0/255 values to 0/1 (1-bit PNG uses 0 for black, 1 for white)
binary_data = (cv_img // 255).tolist()

# Write the 1-bit PNG
with open('output_1bit_pypng.png', 'wb') as f:
    writer = png.Writer(
        width=cv_img.shape[1],
        height=cv_img.shape[0],
        bitdepth=1,
        greyscale=True
    )
    writer.write(f, binary_data)

Quick Checks to Verify #

After saving, you can confirm the bit depth:

• On Windows: Right-click the image → Properties → Details → Bit depth
• On macOS: Right-click → Get Info → More Info → Bits per sample
• Using Pillow: Image.open('output.png').mode returns '1' for 1-bit images.

内容的提问来源于stack exchange，提问作者Jimbo