OCR Test

Experimental app for optical character recognition (OCR).

Runs the Tesseract 3.00 open source OCR engine on images captured by the device camera.

The purpose of this app is to demonstrate OCR running on an Android device. It's interesting to see how the text in some images is recognized better than in others, and how the recognition response time varies with the language of the text being recognized. Conventionally, OCR is run using a flatbed scanner to scan printed pages of text. In contrast, running OCR on images captured by a smartphone/tablet camera on an Android device gives much lower quality--but interesting--recognition results.

The single-shot capture mode runs OCR on a snapshot image that's captured when you click the on-screen shutter button, like a regular photo.

The continuous recognition capture mode loops the recognition so that after one OCR request is completed, another one is immediately started based on whatever is in the camera's field of view at that moment. This mode shows a dynamic, real-time display of what the device is recognizing right beside the camera viewfinder. This mode works best on a fast device.

A translation capability (powered by Google Translate) was added to the single-shot capture mode. The translation results are not really useful on a practical level, though, because the errors in OCR are compounded by the machine translation. Still, it's interesting to try it.