Stamp Image Cropper

I sell stamps on eBay. A single full A4 flatbed scan can contain around 15–20 stamps, and cropping, straightening and saving them as individual images used to take me 10–15 minutes.

It is a necessary part of preparing listings, but it is also repetitive. Every batch involves essentially the same process: find each stamp in the scan, crop it consistently, correct any slight rotation and save it separately.

So I built a Python desktop tool to automate that part of the workflow.

The tool uses OpenCV and Pillow to identify individual stamps placed against a black background, separate them, correct minor rotation and create consistently cropped images.

Under normal conditions, a full A4 scan containing around 15–20 stamps that might take me 10–15 minutes to process manually can now be completed in less than 20 seconds.

That was enough to make the tool worthwhile before anyone else could use it. Saving roughly 10 minutes on a job I repeat regularly adds up quickly.

Turning the desktop tool into a web utility

Once the desktop version was working well in my own workflow, I decided to adapt it into a browser-based version.

The aim was to let other stamp collectors and sellers use the same process without installing Python or image-editing software. I wasn’t trying to invent a different cropping system for the website. I wanted to retain the core process that I was already using.

I treated the desktop tool as the reference implementation. While adapting it, I used reference scans and known-good outputs to check that moving the processing online had not silently changed the cropping behaviour.

The resulting workflow is deliberately simple:

1. Scan several stamps together against a black background.

2. Upload the scan.

3. The cropper identifies and separates the stamps.

4. It corrects minor rotation.

5. It creates the individual JPEG images.

6. It returns them together in a ZIP file.

There are no accounts, no registration and no software to install. The live cropper is here for anyone wanting to try it:

https://www.alstamps.com/image-cropper/

The web architecture

I kept the web implementation relatively small.

The existing Python processing sits behind a Flask API, served by Gunicorn and hosted on Heroku. The public interface is part of my existing WordPress site, but the WordPress server does not handle the image processing.

Broadly, the flow is:

WordPress/browser interface → Python API on Heroku → crop and straighten images → create ZIP → return ZIP to the browser

Temporary upload and output files are deleted automatically after processing.

There is no AI dependency. It is the same core image-processing approach as the desktop tool, made accessible through a browser.

Narrow constraints are fine

The cropper works best when the input follows a few straightforward rules:

- Use a black background.

- Leave visible gaps between the stamps.

- Keep the stamps reasonably square and upright.

- Preferably scan at 600 dpi.

It can handle minor rotation, but it is not intended to straighten stamps placed at arbitrary angles or rescue a badly arranged scan.

That limitation became part of the design rather than something I needed to eliminate. The tool does not have to solve every possible image-segmentation problem. It needs to solve the particular job it was built for reliably.

The other useful lesson for me was that the tool proved its value before it became a public web project. I did not start with plans for a SaaS, user acquisition or a large potential market. I had a repetitive bottleneck in an existing business workflow, built something to remove it and measured a clear time saving.

Only after that did I consider making it available to other people.

I have made the web version free because I originally built the cropper for myself, and there seemed little point keeping it entirely private once I had adapted it for browser use. Other collectors and stamp sellers may have the same repetitive task, and the tool costs them nothing to try.

I’m curious what other indie builders would do with a utility this narrow: keep it as one focused free tool under the existing specialist website, or build a small collection of related tools around the same workflow?