Enhanced ensemble technique for optical character recognition

Abstract

Optical character recognition (OCR) is the electronic transformation of images into a computer-encoded text. OCR systems often produce poor accuracy for noisy images. Ensemble recognition techniques are used to improve OCR accuracy. The idea of the ensemble recognition techniques is to produce N-versions of an input image. These versions are similar but not identical. They are passed through the OCR engine to turn them into different OCR outputs, which later leads to select the best between them. Existing ensemble techniques need to be more effective to reduce OCR error rate. This research proposed enhanced ensemble technique to overcome the drawbacks of existing techniques. The proposed technique was evaluated against three other relevant existing techniques. The performance measurements used in this research were Word Error Rate (WER) and Character Error Rate (CER). Experimental results showed a relative decrease of 14.37% and 40.13% over the WER and CER of the best existing technique. This study contributes to the OCR domain as the proposed technique could facilitate the automatic recognition of documents. Hence, it will lead to a better information extraction.