An Algorithmic Approach for Text Recognition from Printed/Typed Text Images
Authors:
- 1 January 2018
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 876-879
Abstract
Extraction of texts from scanned copies of documents and text images is an important task in the recent scenario. Optical Character Recognition (OCR) is used to analyze text in images. The proposed algorithm deals with taking scanned copy of a document as an input and extract texts from the image into a text format using Otsu's algorithm for segmentation and Hough transform method for skew detection. The system was confined to recognize English alphabets (A-Z, a-z) and numerals (0-9). OCR technique has been implemented to recognize characters. Validation tests were done on screenshots of typed texts and images of scanned document from Internet sources. Experimental results indicate that the proposed algorithm is able to recognize alphabets written in Verdana font style with size 14 and also showed good results with rotated images. The average accuracy to determine rotation angle correctly was calculated to be 90% and overall system accuracy was calculated to be 93%.Keywords
This publication has 5 references indexed in Scilit:
- Optical Character Recognition Program for Images of Printed Text using a Neural Network2006
- Automated segmentation of math-zones from document images2005
- Training on severely degraded text-line images2004
- Handwritten digit recognition: investigation of normalization and feature extraction techniquesPattern Recognition, 2004
- Document image improvement for OCR as a classification problem2003