Graph based features for recognition of handwritten Devanagiri numerals
Mohammad Idrees Bhat, B. Sharada, Panduranga Naidu Nagabhushan · 2016
Automatic recognition of style variant, cursive and irregular unconstrained handwritten devanagiri numerals in scanned images is a challenging task. In this paper, a graph based representation is given to handwritten numerals, as it provides best alternative to overcome the limitations of feature based representation. Besides providing robust representation to handwritten devanagiri numerals, graph based representation is invariant to image transformations (scale, rotation, reflection and translation). Graph dissimilarity space embedding (GDSE) is explored to extract the features. The recognition is carried out with simple nearest neighbor classifier. Extensive experiments have been carried out on CVPR (ISI) Kolkata dataset and the results are quite promising.