Hacker News new | ask | show | jobs
by gravypod 3105 days ago
Will this automatically center and apply perspective transforms to pictures taken with phone cameras?
4 comments

You should check Genius Scan (for iOS and Android) that does that for you automatically, and much more: https://www.thegrizzlylabs.com/genius-scan/

It also doesn't tie you to a specific ecosystem.

Disclaimer: I'm on of the authors of Genius Scan :)

There are many solutions for systems like this. Unfortunately none are open source.

Most documents I do want to keep record of are sensitive. I don't trust a closed-source app with that kind of information.

Microsofts OfficeLen's for iOS is pretty nice doing this automagically.
The best version of this that I've seen is Scanner Pro by Readdle. I had to scan three months worth of food receipts for an insurance claim and this feature was a lifesaver.
seriously, thank you for mentioning this. Great app!
Not at the moment. I'm guessing this information is available in the EXIF properties?
It's a little bit more tricky than that. What the EXIF might tell you is the camera calibration parameters like focal length, distortion, perspective center, etc. That can be used to fix systematic errors in images like pincushion/barrel distortion.

To unwarp photos that were taken at odd angles you need to do some image processing. The mathematics aren't particularly difficult, it's a homography transform in most cases (rectangles). The problem is robustly detecting the page.

Dropbox has some nice write-ups on this: https://blogs.dropbox.com/tech/2016/08/fast-document-rectifi...

Thanks for the link. That blew my mind! I wish it could be added would, my phone would replace my scanner instantly.
Since others are commenting alternative apps, I'll chime in:

https://www.camscanner.com/

I've been using CamScanner on Android since forever. I only stopped a few months ago because I moved to iOS (and didn't realize it's also there).

It does the page detection fairly well, but also let's you correct it. Very useful, specially when the picture is bad. I then used to upload it to Dropbox.

Also see this for more implementation details: https://www.pyimagesearch.com/2014/09/01/build-kick-ass-mobi...

This could be done server side if you're already doing ocr.

Wow this looks awesome, thank you!
Google Drive mobile app does this natively (at least on Android) there's a 'Scan' option when adding a file that will do it.
Scannable (by Evernote ) does this wonderfully. Allows exporting as pdf and storing wherever. I’m a fan.
The most complicated implementation I've seen has been Office Lens, which also corrects distortion and does edge detection.
No. It requires opencv and image processing