如何将OpenCV NumPy数组转为Wand Image格式并保存二值化图像

Hey there! Let's tackle converting that OpenCV-processed NumPy array back to a Wand Image so you can save it properly. Here's how to make it work seamlessly with your existing workflow:

First, let's align with your current pipeline: you're generating a NumPy array from a blob, running it through your binarize_image function to get a binary ndarray, and now need to convert that back to Wand's Image format for saving.

Complete Solution Code #

Here's the adjusted code flow including the critical conversion step:

import cv2
import numpy as np
from wand.image import Image
import os

# Make sure your threshold constant is defined (example value shown)
BINARY_THREHOLD = 127

def binarize_image(img):
    ret1, th1 = cv2.threshold(img, BINARY_THREHOLD, 255, cv2.THRESH_BINARY)
    return th1

# --- Your existing setup ---
# blob = make_blob(...)  # Raw JPEG buffer from your make_blob function
# img_np = np.frombuffer(blob, dtype=np.uint8)
# img_cv = cv2.imdecode(img_np, cv2.IMREAD_GRAYSCALE)  # Load as grayscale

# After running binarization
binary_img_np = binarize_image(img_cv)

# --- Convert OpenCV ndarray to Wand Image ---
with Image(
    blob=binary_img_np.tobytes(),
    width=binary_img_np.shape[1],
    height=binary_img_np.shape[0],
    format='gray'  # Specify single-channel grayscale format
) as wand_img:
    # Save using your original path logic
    output_path = os.path.join(pdf_folder, image_folder, outputFileName)
    wand_img.save(filename=output_path)

Key Details to Note #

• NumPy to Raw Bytes: Your binary ndarray is a 2D array of uint8 values (0 or 255 for binary images). Using .tobytes() converts this array into a raw byte stream that Wand can parse directly.
• Wand Image Setup: We pass the raw bytes via the blob parameter, then set the exact dimensions (remember: OpenCV uses (height, width) for array shape, so we flip those for Wand's width/height params). The format='gray' flag tells Wand this is a single-channel grayscale image, which matches your binary output.
• Context Manager: Using with ensures the Wand Image resource is cleaned up automatically after saving, which is a good practice to avoid memory leaks.

Quick Validation Checks #

• Confirm your binary_img_np is a 2D array (no extra color channels). If you loaded the original image as BGR (3-channel), cv2.threshold still works, but double-check to avoid unexpected formatting issues.
• Ensure BINARY_THREHOLD is properly defined in your code (the example value above is just a placeholder—use your actual threshold value).

That should get your image saved correctly in Wand's format without a hitch! 😊

内容的提问来源于stack exchange，提问作者Anee