The AI-900 exam has been retired
It was retired on June 30, 2026. It was replaced by AI-901 (Microsoft Azure AI Fundamentals). Go to the AI-901 study guide →
4.1.3. Optical Character Recognition (OCR)
OCR extracts text from images. It converts visual text (in photos, scanned documents) into machine-readable text.
Key characteristics:
- Input: Image containing text
- Output: Extracted text strings with position information
- Works on printed and handwritten text
How OCR works:
- Text detection: Find regions containing text in the image
- Text recognition: Convert detected regions to character strings
- Layout analysis: Understand reading order and structure
- Output generation: Return text with bounding boxes
OCR output includes:
- Extracted text content
- Bounding box coordinates for each text region
- Confidence scores for recognition accuracy
- Reading order inference for multi-column layouts
Common scenarios:
- Digitizing paper documents
- Reading text from signs
- Extracting data from receipts
- Processing medical records
- Capturing business card information
- Accessibility applications (text-to-speech from images)
OCR vs Document Intelligence:
| Capability | OCR | Document Intelligence |
|---|---|---|
| Primary function | Extract raw text | Extract structured data |
| Output | Text strings | Fields, tables, key-value pairs |
| Use case | General text extraction | Form processing, invoices |
⚠️ Critical Exam Trap: OCR is a COMPUTER VISION workload, not NLP! The input is an image. The fact that the output is text doesn't change this—always classify by input type.