Most OCR (Optical Character Recognition) systems developed to recognize texts embedded in multimedia documents segment the text into characters before recognizing them. In this paper, we propose a novel approach able to avoid any explicit character segmentation. Using a multi-scale scanning scheme, texts extracted from videos are first represented by sequences of learnt features. Obtained representations are then used to feed a connectionist recurrent model specifically designed to take into account dependencies between successive learnt features and to recognize texts. The proposed video OCR evaluated on a database of TV news videos achieves very high recognition rates. Experiments also demonstrate that, for our recognition task, learnt feature representations perform better than hand-crafted features.
展开▼
