I have a following question. I would like to read these types of captcha in python:

The best code I have done is this, however it is not able to solve all these captchas:

import pytesseract
import cv2
import numpy as np
import re

def odstran_sum(img,threshold):
    """Funkce odstrani sum."""
    filtered_img = np.zeros_like(img)
    labels,stats= cv2.connectedComponentsWithStats(img.astype(np.uint8),connectivity=8)[1:3]
    label_areas = stats[1:, cv2.CC_STAT_AREA]
    for i,label_area in enumerate(label_areas):
        if label_area > threshold:
            filtered_img[labels==i 1] = 1
    return filtered_img


def preprocess(img_path):
    """Konvertuje do binary obrazku."""
    img = cv2.imread(img_path,0)
    blur = cv2.GaussianBlur(img, (3,3), 0)
    thresh = cv2.threshold(blur, 150, 255, cv2.THRESH_BINARY_INV)[1]
    filtered_img = 255-odstran_sum(thresh,20)*255
    kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (3,3))
    erosion = cv2.erode(filtered_img,kernel,iterations = 1)
    return erosion

def captcha_to_string(obrazek):
    """Funkce vrati text z captchy"""
    text = pytesseract.image_to_string(obrazek)
    return re.sub(r'[^\x00-\x7F] ',' ', text).strip()

img = preprocess(CAPTCHA_NAME)
text = captcha_to_string(img)
print(text)

Is it possible to improve my code that it will be able to read all these five examples? Thanks a lot.

CodePudding user response：

I don't think there is much to be improved beside writing own neural network for image recognition based on similar captchas. Captchas are rather designed so that computer has hard time decoding them, so I don't think you can get perfect results.