Rectification of camera captured document images for camera-based OCR technology

摘要

In this work we propose a fully automatic pre-processing technique to enhance the digital camera captured images and rectify the different known types of distortion to improve the performance of OCR applications. Our proposed approach depends on the features of the text lines and letters and doesn't need any especial equipment. Experimental results on a real mobile digital camera (Samsung Galaxy S2 with 8-MP sensor) captured images demonstrate 10~16 % enhancement in current commercial Arabic OCR packages.