Top-down and bottom-up cues for scene text recognition | IEEE Conference Publication | IEEE Xplore