I have millions of pictures of things (buildings, animals, people, etc).
Some of the pictures are of bad quality, so they need to be shot again.
How to measure the "quality" of each picture?

I define "quality" as:

  • Size: the bigger the better. Above 2000*2000 pixels is already good.
  • Contrast: Colors should go from black to white, a picture of a building taken in sunny weather is better than a picture taken in foggy weather.
  • Sharpness: The picture should not be blurry nor oversampled (too many pixels compared to the actual detail level of the lens/camera)
  • Shape: Height and width should be of the same magnitude. A very narrow strip is not a quality picture for me.

The overall score should be a combination of these, a very bad score in one criteria should result in a bad overall score, even if the picture is ultra-good for one other criteria.


  • Open source
  • Runs on Linux
  • The faster the better (for the same level of accuracy)

There will be false positives, for instance even a great picture of an Aurora Borealis will probably be considered too blurry. A picture of the Milky Way will probably be considered too dark, unless it is heavily edited. A picture of fog will be considered too foggy. These kinds of false positives are acceptable.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.