నైరూప్య

View Based Feature Extraction and Classification Approach to Malayalam Palm Leaf Document Image

Geena K.P, Raju. G

Malayalam Handwritten Character recognition is still an active area of research. Most of the contemporary works reported use artificial data set taking only 44 characters. Up to 99.78% accuracy is reported with 450 samples per character. In this paper we focus on the recognition of characters extracted from Palm leaf (PL)manuscripts. Unlike the synthetic dataset used PL manuscript images pose more challenges in the pre-processing and recognition stage. To study the performance of the existing Handwritten Character Recognition (HCR) system on PL images, we created a database consisting of 450 samples each of 44 chosen Malayalam character obtained from PL images. For the purpose of comparison a synthetic database consisting of 450 samples each of the same 44 characters are used.