Using SIFT Descriptors for OCR of Printed Arabic

Although optical character recognition of printed texts has been a focus of research for the last few decades, Arabic printed text, being cursive, still poses a challenge. The challenge is twofold: segmenting words into letters and identifying individual letters. We propose a method that combines the two tasks, using multiple grids of SIFT descriptors as features. To construct a classifier, we don’t use a large training set of images with corresponding ground truth, a process usually done to construct a classifier, but, rather, an image containing all possible symbols is created and a classifier is constructed by extracting the features of each symbol. To recognize the text inside an image, the image is split into “pieces of Arabic words” (paws), and each paw is scanned with increasing window sizes. Segmentation points are set where the classifier achieves maximal confidence. Using the fact that Arabic has four forms of letters (isolated, initial, medial and final), we narrow the search space based on the location inside the paw. The performance of the proposed method, when applied to printed texts and computer fonts of different sizes, was evaluated on two independent benchmarks, PATS and APTI. Our algorithm outperformed that of the creator of PATS on five out of eight fonts, achieving character correctness of 98.87%-100%. On the APTI dataset, ours was competitive or better that the competition.