Text this: Fractional means based method for multi-oriented keyword spotting in video/scene/license plate images