Multiwavelet and connected pixel based feature for handwritten Marathi characters
Ravindra S. Hegadi, Parshuram M. Kamble, Anil S. Sherikar, B.V. Dhandra · AIP conference proceedings · 2018
Recognition of handwritten Marathi characters is challenging field in image processing and character recognition system. In this proposed method, we extract futures from handwritten Marathi characters using multiwavelet and connected pixel based feature extraction methods. Classification of handwritten Marathi characters is done by using the bagged tree and Quadratics discriminate classifier. Before feature extraction and classification, handwritten character images were enhanced using preprocessing techniques. Experimental analysis and performance of proposed methods were measured using five-fold cross-validation. The proposed model reduces the false acceptance during recognition of handwritten Marathi characters.