A text detection, localization and segmentation system for OCR in images

One way to include semantic knowledge into the process of indexing databases of digital images is to use caption text, since it provides important information about the image content and is a very good entity for queries based on keywords. In this paper, we propose an approach to automatically localize, segment and binarize text appearing in complex images. First, an unsupervised method based on a wavelet transform is used to efficiently detect text regions. Second, connected components are generated, and the exact text positions are found via a refinement algorithm. Third, an unsupervised learning method for text segmentation and binarization is applied using a color quantizer and a wavelet transform. Comparative experimental results demonstrate the performance of our approach for the main processing steps: text localization and segmentation, and in particular their combination.