Programs
- M. Tech. in Automotive Engineering -
- Fellowship In Interventional Pulmonology and Pulmonary Critical care - Fellowship
Publication Type : Conference Paper
Publisher : 2017 International Conference on Intelligent Computing and Control
Source : 2017 International Conference on Intelligent Computing and Control (I2C2), IEEE (2018)
Url : https://ieeexplore.ieee.org/abstract/document/8321962
Keywords : distance profiles, English OCR, font style classification, printed character recognition, SVM classifier
Campus : Mysuru
School : School of Arts and Sciences
Department : Computer Science
Abstract : The inclination of optical technologies like OCR lies in achieving higher recognition rates with optimal or reduced computational complexities. At present there exist optical technologies for automation of reading the text from document images with almost nearing to 100% accuracy. Especially, the Roman language OCR's are reliable and robust enough in producing higher accuracies by being able to recognize varying font styles of varying sizes. However for the font style/ size independent OCR's one of the main aspect is its computational complexity. It is significant concern to reduce the computational complexities involved in the process of character recognition through a font style / size independent OCR. In this paper, a technique for classification of the font style based on character image is proposed by employing the distance profile features with respect to left, right and diagonal directions of a character image. The major objective of this work is to reduce the complexity of the generic OCR systems by font style recognition. The distance profile features of character images are fed to a support vector machine classifier. For experimentation, the training data sets are comprised of around 10 widely used font styles of upper case letters as well as lower case letters. The testing is conducted with the character images that are extracted from various non editable document sources comprising of 5 different font styles. The performance of algorithm is found to be satisfactory with an accuracy of 80%. © 2017 IEEE.
Cite this Research Publication : Bharath V. and Shobha Rani N., “A font style classification system for English OCR”, in 2017 International Conference on Intelligent Computing and Control (I2C2), 2018.