Improving State-Of-The-Art OCR Through High-Precision Document-Specific Modeling

doi 10.1109/cvpr.2010.5539867