使用Python去除图片验证码干扰线的技术实现问询

Hey there, I tweaked existing captcha line removal code to target the interference lines in your test captcha. Here's the modified implementation ready to use:

from PIL import Image, ImageFilter
from scipy.misc import toimage
from operator import itemgetter
from skimage import measure
import numpy as np
import heapq
import cv2
import matplotlib.pyplot as plt
from scipy.ndimage.filters import median_filter

#----------------------------------------------------------------
class Preprocess:
    def __init__(self):
        pass

    def remove_interference_lines(self, img_path):
        # Load grayscale captcha image
        raw_img = cv2.imread(img_path, 0)
        # Apply median filter to reduce noise without blurring characters
        filtered_img = median_filter(raw_img, size=3)
        
        # Use connected component analysis to identify regions
        labels = measure.label(filtered_img, connectivity=2)
        region_props = measure.regionprops(labels)
        
        # Filter out regions that match interference line characteristics (thin & long)
        line_regions = [
            prop for prop in region_props 
            if (prop.bbox[3] - prop.bbox[1] < 6) and (prop.bbox[2] - prop.bbox[0] > 25)
        ]
        
        # Mask out identified line regions
        for region in line_regions:
            min_row, min_col, max_row, max_col = region.bbox
            filtered_img[min_row:max_row, min_col:max_col] = 255
        
        # Final smoothing to clean up edges
        final_clean_img = Image.fromarray(filtered_img).filter(ImageFilter.SMOOTH)
        return final_clean_img

# Example usage to test your captcha
if __name__ == "__main__":
    processor = Preprocess()
    cleaned_captcha = processor.remove_interference_lines("your_test_captcha.png")
    cleaned_captcha.save("cleaned_captcha_output.png")
    
    # Display the result
    plt.imshow(cleaned_captcha, cmap='gray')
    plt.axis('off')
    plt.show()

Key Changes I Made: #

• Added median filtering to suppress noise while keeping character edges sharp
• Used connected component analysis to distinguish thin, elongated interference lines from captcha characters
• Built a region-filtering rule based on aspect ratio to target only line-like areas
• Added post-processing smoothing to refine the final cleaned image

内容的提问来源于stack exchange，提问作者rut_0_1