New model method for distorted document image correction
Wang Jing-zhon · Information technology newsletter · 2015
As a result of book pages bending,document images taken by camera may be distorted to some degree and not proper for optical character recognition( OCR) software to recognize. A new methemetical model is proposed to solve the problem. First,text lines are iterated through to find points that mark words postion. Then,position points are fitted according to the model to find optimum curves.Finally,text pixels are recovered according to the differece between fitting curve and horizontal line. The experiment results show that distorted document images are correctly recovered with good time efficiency.