Arabic OCR Based on Image Invariants

Abstract

Optical character recognition (OCR) is an area of interest for converting hard copies of written information, as in faxed or scanned documents. In this paper, we compare two common image invariant methods in OCR, the invariant central moments (NCM) and the normalized Fourier descriptors (NFD) of the character. These methods are insensitive to image variances such as shifting, scaling, or rotation. To ensure the efficiency of the proposed algorithms, some examples and simulation are done. We also show by means of examples that when using the NCM the recognition rate decreases when the image coordinates ratio is changed