Digitising Bank Cheque Processing with Advanced OCR API Solutions

The Azimut Bank Cheque OCR API gives developers a single endpoint for extracting cheque data — bank name, cheque number, date, payee, amounts in words and numerals — from scanned cheque images. Whether you are integrating cheque parsing into a mobile banking app, a back-office system, or a fraud detection pipeline, the API returns structured JSON ready for validation and posting.

The API is modular by design. It supports real-time single-cheque validation, batch processing for high-volume workflows, and custom data handling rules per deployment.

How it works

Submit a base64-encoded cheque image to the API endpoint. The extraction model reads the image, identifies each field region, and returns structured data — bank name, branch, cheque number, date, payee, amount in words, and amount in numerals — with confidence scores per field.

Using cURL:

1
2
3
curl -X POST "https://api.azimut.com/cheque/parse" \
  -H "Authorisation: Bearer YOUR_API_KEY" \
  -d '{"cheque_image": "BASE64_ENCODED_IMAGE"}'

Using Postman, send a POST request to

with an header, , and a JSON body containing the base64-encoded image.

The response is a structured JSON payload:

1
2
3
4
5
6
7
8
9
10
11
{
  "content": {
    "Bank Name": "XYZ Bank",
    "Branch": "Main Branch",
    "Cheque Number": "123456",
    "Date": "2025-02-04",
    "Payee": "John Doe",
    "Amount in Words": "Five Thousand Dollars Only",
    "Amount in Numerals": "5,000"
  }
}

The data is ready for downstream processing — posting to a core banking system, running fraud checks, or feeding into a batch clearing file. The same output format works whether the cheque came from a self-service kiosk scanner, a teller's desktop scanner, or a mobile phone photo.

Where it applies in production

Automated cheque deposits. Banks integrate the API into mobile and web banking channels so customers submit cheques through the app. The API extracts and validates the data, and the deposit posts without manual handling.

Fraud detection. The extracted fields feed into analysis workflows: amount-in-words versus numerals, date staleness, duplicate serial checks, and watchlist screening. Banks layer their own fraud rules on top.

Back-office automation. Finance teams processing large volumes of incoming cheques send scanned images through the API and receive structured data without manual key-entry.

Batch processing. For high-volume scenarios, the API accepts multiple images and returns structured data for each cheque in a single batch response.

The same extraction engine handles printed and handwritten fields. Handwritten payee names, amounts in words, and dates — the fields that cause the most errors in manual processing — are read by the ICR models trained on real cheque datasets. Image quality matters; higher-resolution scans produce better results.

Security and compliance

OCR data processing stays within the bank's infrastructure. The API runs in the bank's own deployment environment — on-premise or cloud — and the extracted data never leaves the bank's control. Compliance with cheque truncation regulations and local clearing rules is built into the validation layer, not added as an afterthought.

---

To see the extraction pipeline in production — MICR validation, signature verification, clearing integration — see the Cheque OCR & Fraud Detection use case. For the full picture of cheque field extraction at 97%+ accuracy, read how the pipeline reaches field-level accuracy on handwritten cheques.