Structural Features Extraction for Devnagari and Bangla Language Documents

Abstract

India is a multi-lingual, multi script country. Therefore developing a successful multi-lingual OCR, system for feature extraction of different scripts is a very important step. In this paper we discussed a Structural Features based algorithm for feature extraction. A family of procedures for measuring relevant shape information in a pattern in order to make the task of classifying pattern easy is called Feature Extraction. It analyses a segment of text and selects features which are unique to the text and can be used to identify it. The soul of pattern recognition system design is selection of features that are stable and representative. The real objective of feature extraction is to concentrate a situated of features, which expands the distinguishment rate with the slightest measure of components and to produce comparable feature set for mixture of cases of the same image. Feature extraction systems investigate the data archive picture and select a set of features that extraordinarily distinguishes and characterizes the character. Feature Extraction and Classification procedures are essential steps in character distinguishment methodology to attain to high distinguishment execution. Feature extraction is characterized as the issue of “separating from the crude information the data which is most important for characterization purposes, in the feeling of minimizing the inside class design variability while upgrading the between-class design variability”. In this paper we have developed structural features for Devnagari and Bangla script document.