How to Crack Image CAPTCHAs with Python: From Noise Reduction to OCR

Recognizing image CAPTCHAs is a fundamental skill for web crawling and involves computer graphics, machine learning, computer vision, and artificial intelligence.

The core of computer graphics is representing and processing geometric elements (points, lines, surfaces) and non‑geometric attributes (color, shading). CAPTCHA cracking mainly uses 2‑D elements and color‑difference analysis.

Common tools include:

Support Vector Machine (SVM)

OpenCV

Image editing software (Photoshop, GIMP)

Python Imaging Library (PIL)

PIL installation

On Debian/Ubuntu:

<code>sudo apt-get install python-imaging</code>

On other Linux distributions you can use easy_install PIL or pip install Pillow after installing the build environment. For Windows download the installer from the official PIL site (32‑bit) or the 64‑bit Pillow wheels from Gohlke .

General CAPTCHA recognition workflow

Image denoising

Image segmentation (cutting)

Text extraction

1. Denoising

Convert the color image to grayscale (using the intensity component of the HSI color space):

<code>from PIL import Image im = Image.open('7039.jpg') imgry = im.convert('L') imgry.show()</code>

2. Binarization

Apply a fixed threshold (e.g., 140) to create a binary image:

<code>threshold = 140 table = [] for i in range(256): if i < threshold: table.append(0) else: table.append(1) out = imgry.point(table, '1') out.show()</code>

3. Segmentation

Segmentation is the most challenging step, especially for tightly connected characters; simple CAPTCHAs can be processed without complex cutting. For advanced methods see related blogs.

4. OCR with pytesser

pytesser is a Python wrapper for Google’s Tesseract OCR engine.

Installation steps:

Download and install PIL (or Pillow) if not present.

Download pytesser, extract it into your project or Python’s site‑packages directory, and add it to PYTHONPATH.

Download the Tesseract OCR engine, replace the tessdata folder inside pytesser with the one from the engine.

Modify pytesser.py to import Image from PIL and add an empty __init__.py in the pytesser package.

Example usage:

<code>from PIL import Image from pytesser import pytesser image = Image.open('7039.jpg') print(pytesser.image_file_to_string('7039.jpg')) print(pytesser.image_to_string(image))</code>

pytesser also supports other languages, including Chinese.

Source: j_hao104 – my.oschina.net/jhao104/blog/647326