Discrete Cosine Transform for Script Identification and Character Recognition

Optical Character Recognition (OCR) in printed multi-script documents is still challenge due to the script dependence of OCR. Identification of script is important phase in design of multi-script OCR system for processing of multi-script documents. Most of the script identification work reported is on document, paragraph/block, and word level. This research article presents character level script identification and character recognition using Discrete Cosine Transforms (DCT) feature in bilingual Gujarati-English text. DCT is employed to extract the features based on energy coefficients analysis. The proposed method has two phases: Classification and Recognition. In classification, performance of KNN and SVM classifiers is studied separately and compared. The same DCT features are used in recognition phase. Experiments and results show that, presented method is robust for character level printed bilingual script identification and character recognition.