Reputation: 15780
I'm trying to develop and algorithm to read cursive handwriting (doctors' handwriting to be exact). I don't need to match every single letter, just the recognizable ones (I'm using dictionary approach).
For now, I have the binarized image in BufferedImage
class. What I need to do next is to isolate each character.
Can you provide me of a pseudocode on how can I achieve this? I would like to be minimal in using other libraries. I want to do this by learning how each process is achieved. Any hints?
Upvotes: 0
Views: 2841
Reputation: 6750
You will need to learn about stuff such as Image Moments that can be used to compare images even if they have been rotated or scaled
You would need heavy math background to understand how to do that, you can start by reading this http://www.ias.ac.in/sadhana/Pdf2007Oct/521 , it is about how an OCR for a particular kind of characters (Kannada) was implemented, the same general principles should apply for what you are trying to do
Unless you are willing to spend several months (or years) on this pursuit I would recommend you to use something already available, like ocropus or javaocr
Upvotes: 2
Reputation: 11
Alternatively you can purchase a HWR SDK from Vision Objects that will considerably reduce your development effort.
Upvotes: 1