Enhancing Pre-Printed Character Recognition in Noisy Environments using Back Propagation Neural Networks

Abstract

The original picture and its characteristics are pre-processed using a number of techniques including noise filtering, smoothing, thresholding, and skewing in order to increase OCR accuracy in this article. Experiments with many kinds of noisy and non-uniform character documents demonstrate that recognition side enhancement has been successful. This study reports the results of efficiency and recognition tests conducted on a training approach. This research demonstrates how to utilize an optical character recognition application with an artificial neural network to achieve top performance and recognition quality using several image processing techniques. Data may be easily updated, processed, and saved in a computer-readable format once it has been converted to digital form using an OCR (Optical Character Recognition) system.