Segmentation and Recognition of Bengali Characters from Document Images

In this paper, a method to segment the Bengali character from document images and recognition of each character in multiple categories has been proposed. The proposed algorithm works in two stages. Firstly, characters are segmented from the document page from horizontal and vertical projection profile. Secondly, characters are recognized by a support vector machine (SVM) classifier using different features extracted from the characters. The proposed methodology has been evaluated by a number of document images and the experimental results demonstrate the efficiency of the proposed method.