1

I am trying to segment my characters on this image with using column sum vector.

This method actually works on the paper I've read about license plate recognition.

it looks useful for my images too but since my threshold values are not very good I am not really sure if I can use CSV for character segmentation

enter image description here enter image description here

so my question is;

1) Is CSV proper method for my application ? 2) if so what kind of techniques I can try to segment characters ? 3) if not what can be more efficient and robust way to segment characters from some kind of images ?

thank you

Anar Bayramov
  • 257
  • 1
  • 4
  • This question is very related to : http://dsp.stackexchange.com/questions/19793/suggested-preprocessing-methods-for-ocr-on-circular-images/19794#19794

    and :

    http://dsp.stackexchange.com/questions/19837/feature-detection-on-a-circular-metal-surface-for-character-recognition/19838#19838

    I guess the other threads could be used and this is not necessarily a very distant question.

     – Tolga Birdal May 01 '16 at 18:55

1 Answers1

1

Your results (above) looks well for character segmentation. But if you have some problems with calculation of thresholds for total image, you can try slightly different approach. "Column sum vector" method works well for grey images too, but not only for black/white images. So there is no need to calculate threshold for total image, but you must choose some value for segmentation in vector. May be it will slightly simpler to do segmentation in this case.

You can use "line sum vector" for vertical segmentation too.

SergV
  • 930
  • 5
  • 10