Gabor Filter Based Technique for Offline Indic Script Identification from Handwritten Document Images

Script identification from a document image is a complex real life problem in a multi-script country like India. The work becomes more challenging when handwritten documents are considered. In this paper, a Gabor filter based technique has been developed for offline script identification from handwritten document images. The work is carried out at document level on four popular Indic scripts namely Bangla, Devanagari, Roman and Urdu. A total of 157 handwritten document images are considered from these four scripts for experimentation. The data set is divided into training and test set in 2:1 ratio. A feature vector of 20 dimensions is constructed using Gabor filter and Morphological reconstruction. Finally, using MLP classifier, a recognition accuracy of 95.4% is obtained on test data without any rejection.