Thanks for the answer...
the "A" is only an example caracter.
in the real world example i use japanese caracters.
the user shall be paint the correct caracter in my app, like this on the picture...
That will be an insanely difficult task. Actually this in general takes years of study in order to accomplish
Handwriting recognition is one of the hardest things to accomplish.
If you do want to give it a shot, my suggestion:
Crop and rescale the images, and than determine patterns for each letter, thus an A can be build from 3 linear formules, check if the drawing matches this structure. You can than compute the derivatives of the drawings and from those derivatives cross check them with a database to determine which letter it is.
But this is extremely difficult, we tried to read digits in a sudoku puzzle which was already quite a difficult task to accomplish (and we tried to reference it against a database with images, as well as checking several characteristic points in a figure etc) this went OK with printed letters, but with handwritten it was a disaster. Not trying to discourage you, maybe there are libraries out there which you can use, but I would reconsider what you are trying to accomplish and determine an approach for yourself.