This project aims to develop a MATLAB algorithm that flattens any warped image containing text. The goal is to make the text in original image more readable for humans and OCR (optical character recognition).
Executing MATLAB code requires an installation of MATLAB. Open source alternatives such as GNU Octave may also work. To execute the dewarping algorithm on the sample image, run:
script.m
The current algorithm improved the OCR accuracy from 42% in original image to 87% in dewarped image. Processing time was ~30 seconds for the sample image. More in depth description of this algorithm's development is in project summary.
Current algorithm work only with mostly upright images. It can be improved by the addition of a detection algorithm to make rotated images upright. More advanced image processing functionality could be achieved by integrating libraries like OpenCV into MATLAB.