We suggest a system for recognizing handwriting text from a whiteboard. In contrast to the mainstream approaches, we are able to recognize text solely from still images and do not require additional hardware or tracking of the writer’s hand. The proposed system contains extraction of text areas, suitable image preprocessing steps, line extraction, and finally text recognition. Since we are dealing with realistic, and thus very difficult data, a special emphasize of this contribution lies on the preprocessing steps. Experimental results on a large, challenging benchmark set clearly justify further investigations of the proposed approach. Despite the difficulty of the benchmark data used, we come close to a reference approach operating on rendered, near perfect synthetic data.