A comparative analysis of image transformations for handwritten Odia numeral recognition

The work proposed in this paper is an attempt to develop two recognizers for Odia handwriting based on basic transformation schemes and compares their pros and cons. Both the recognizers put emphasis on exploiting the inherent characteristics of the Odia numeral images. The proposed method analyzes the use of Discrete Cosine Transformation (DCT) and Discrete Wavelet Transformation (DWT) for this purpose. Recognition by classification is achieved by feeding these vectors as input to a Back Propagation Neural Network (BPNN). Recognition results are obtained and compared by experimentally varying the classifier parameters. Our experimental results come out to be promising. Thus, finally, we come of with a robust recognizer for handwritten Odia numerals.