Font Acknowledgment and Character Extraction of Digital and Scanned Images

TL;DR

This paper presents a correlation-based method for recognizing fonts and extracting characters from digital and scanned images, even with noise, useful for processing surveillance footage and similar data.

Contribution

It introduces a simple correlation-based approach for font recognition and character extraction from noisy images, implemented in MATLAB.

Findings

Effective font recognition for Times New Roman, Arial, Comic Sans MS

Successful character extraction with noise resilience

Applicable to surveillance and monitoring footage

Abstract

The font recognition and character extraction is of immense importance as these are many scenarios where data are in such a form, which cannot be processed like in image form or as a hard copy. So the procedure developed in this paper is basically related to identifying the font (Times New Roman, Arial and Comic Sans MS) and afterwards recovering the text using simple correlation based method where the binary templates are correlated to the input image text characters. All of this extraction is done in the presence of a little noise as images may have noisy patterns due to photocopying. The significance of this method exists in extraction of data from various monitoring (Surveillance) camera footages or even more. The method is developed on Matlab\c{opyright} which takes input image and recovers text and font information from it in a text file.