Reading between lines: high-rate, non-intrusive visual codes within regular videos via ImplicitCode

Given the penetration of mobile devices equipped with cameras, there has been increasing interest in enabling user interaction via visual codes. Simple examples like QR Codes abound. Since many codes like QR Codes are visually intrusive, various mechanisms have been explored to design visual codes that can be hidden inside regular images or videos, though the capacity of these codes remains low to ensure invisibility. We argue, however, that high capacity while maintaining invisibility would enable a vast range of applications that embed rich contextual information in video screens. To this end, we propose ImplicitCode, a high-rate visual codes that can be hidden inside regular videos. Our scheme combines existing techniques to achieve invisibility. However, we show that these techniques, when employed individually, are too constraining to deliver a high capacity. Experiment results show that ImplicitCode can deliver a significant capacity boost over two recent schemes, up to 12x that of HiLight [19] and 6x or 7x that of InFrame [32], while maintaining a similar or better level of invisibility.