Effect of the perturbed correlation method for optical character recognition

摘要

In order to enhance the ability of correlation method in character recognition, we introduce the perturbation where all six parameters of a two-dimensional affine transformation are changed slightly from the original value, the correlation for all cases of the parameter set are tried, and the best correlation is taken as the final result. We used reciprocal directional features as the feature representation for the correlation, and it was applied to the handwritten Kanji character database ETL8b2. As a result of the experiment, the recognition accuracy was improved by 5%, and the maximum correct recognition rate was 97.4%. The problem to be solved is the enormous amount of computation that the procedure requires.