OCR for Scanned Documents and Images

When You Need OCR

Digital PDFs have extractable text. Scanned documents don't - they're just images.

OCR converts:

• Scanned contracts
• Old books
• Receipts
• Screenshots
• Handwritten notes

Tesseract (Free, Open-Source)

DEVELOPERpython
from PIL import Image
import pytesseract

def ocr_image(image_path):
    image = Image.open(image_path)
    text = pytesseract.image_to_string(image, lang='eng')
    return text

# With confidence scores
data = pytesseract.image_to_data(image, output_type=pytesseract.Output.DICT)
for i, word in enumerate(data['text']):
    confidence = data['conf'][i]
    if confidence > 60:  # Filter low-confidence
        print(word)

Preprocessing for Better Accuracy

DEVELOPERpython
import cv2
import numpy as np

def preprocess_image(image_path):
    img = cv2.imread(image_path)
    
    # Convert to grayscale
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    
    # Remove noise
    denoised = cv2.fastNlMeansDenoising(gray)
    
    # Threshold (binarize)
    _, binary = cv2.threshold(denoised, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)
    
    return binary

# Use preprocessed image
preprocessed = preprocess_image('scan.jpg')
text = pytesseract.image_to_string(preprocessed)

AWS Textract (Commercial, Best Quality)

DEVELOPERpython
import boto3

textract = boto3.client('textract', region_name='us-east-1')

def extract_with_textract(image_path):
    with open(image_path, 'rb') as f:
        image_bytes = f.read()
    
    response = textract.detect_document_text(
        Document={'Bytes': image_bytes}
    )
    
    text = ""
    for block in response['Blocks']:
        if block['BlockType'] == 'LINE':
            text += block['Text'] + '\n'
    
    return text

# For tables and forms
response = textract.analyze_document(
    Document={'Bytes': image_bytes},
    FeatureTypes=['TABLES', 'FORMS']
)

Google Cloud Vision

DEVELOPERpython
from google.cloud import vision

client = vision.ImageAnnotatorClient()

def ocr_with_google(image_path):
    with open(image_path, 'rb') as f:
        content = f.read()
    
    image = vision.Image(content=content)
    response = client.document_text_detection(image=image)
    
    return response.full_text_annotation.text

Handwriting Recognition

DEVELOPERpython
# EasyOCR for handwriting
import easyocr

reader = easyocr.Reader(['en'])
result = reader.readtext('handwritten.jpg', detail=0)
text = ' '.join(result)

Multilingual OCR

DEVELOPERpython
# Tesseract with multiple languages
text = pytesseract.image_to_string(
    image,
    lang='eng+fra+deu'  # English + French + German
)

# EasyOCR (better for Asian languages)
reader = easyocr.Reader(['en', 'zh', 'ja', 'ko'])
results = reader.readtext('multilingual.jpg')

OCR opens up scanned content for RAG. Essential for legal, healthcare, and historical documents.