A novel scheme for Bengali handwriting recognition based on morphological operations with adaptive auto-generated structuring elements

This paper proposes a novel scheme for recognising Bengali handwritten numerals using mathematical morphology. The numerals are broadly classified into two groups based on the presence and position of blobs and stems in them. Since different writing styles are used by different persons, morphological operations with the same structuring elements (SEs) do not yield satisfactory result. Thus, this paper proposes a scheme for automatic generation of SEs based on common handwriting styles. Also the SEs are scaled automatically according to the size of the handwritten digits in order to make the algorithm more robust and efficient. Since, this method does not require any kind of `learning' to work, it is considerably faster than other machine learning based algorithms. Also, it does not require training samples. The present algorithm is tested on a large database of Bengali handwritten digits and its performance and accuracy are also determined.