Extract letter areas from the pictures of text
You would need to take the image from a list and store the coordinates (top left, bottom right) of each area in which there is a text to be found and for some images where there is a letter/character to be found.
You have to write these coordinates in a text file:
for each file in the list write:
- for each letter (and punctuation marks like" ' " or "!") : (top-left pixel, bottom-right pixel)
- text area: (top-left pixel, bottom right pixel)
* punctuation marks should be named as per its unicode character number from Basic Latin http://en.wikipedia.org/wiki/List_
If you work using bbtesseract or http://code.google.com/p/imageclip
I may also ask you to supply cleaned 28x28 pixels images with the background to the character made uniform for a subset of the images.