Extracting Low Resolution Text with an Active Camera for OCR

Reading text in any scene is useful in the context of wearable computing, robotic vision or as an aid for visually handicapped people. Here, we present a novel automatic text reading system using an active camera focused on text regions already located in the scene (using our recent work). A region of text found is analysed to determine the optimal zoom that would foveate onto it. Then a number of images are captured over the text region to reconstruct a high-resolution mosaic of the whole region. This magni ed image of the text is good enough for reading by humans or for recognition by OCR. Even with a low resolution camera we obtained very good results.