Mobile Reader: Turkish scene text reader for the visually impaired

Reading text is one of the crucial needs of the visually impaired people. In this work, we developed a mobile system (Mobile Reader) that can read aloud Turkish scene and book text. Images acquired from the camera are processed in realtime to detect text regions, recognize the text and synthesize speech in Turkish. A fast gradient-based text detection algorithm is employed to enable real-time operation, and character recognition is achieved via the open source Tesseract OCR engine. The performance of the system (running time and accuracy) is measured on a scene text image dataset on a mobile device.