OCR for Arabic using SIFT Descriptors With Online Failure Prediction

Character recognition for Arabic texts poses a twofold challenge, segmenting words into letters and identifying the individual letters. We propose a method that combines the two tasks, using a grid of SIFT descriptors as features for classification of letters. Each word is scanned with increasing window sizes; segmentation points are set where the classifier achieves maximal confidence. Using the fact that Arabic has four types of letters, isolated, initial, middle and final, we are also able to predict if a word is correctly segmented. Performance of the algorithm applied to printed texts and computer fonts was evaluated on the PATS-A01 dataset. For fonts with non-overlapping letters, we achieve letter correctness of 87–96% and word correctness of 74–88%. For overlapping fonts, although the word correctness is low, only 14–23% are not predicted to be wrong. We suggest several approaches for improved performance, with and without exploiting failure prediction. Keywords-OCR, Optical, Character Recognition, Arabic, SIFT, Machine Learning, Predicting Correctness