Object recognition for use in point-and-link communication

Device-to-device (D2D) communication has been becoming increasingly popular with the spread of mobile terminals including tablets and smart phones. D2D communication has received considerable attention due to its benefits of offloading cellular traffic and creating new applications such as location-aware services and content distribution. Recently, a new type of D2D communication, referred as “point-and-link communications”, was proposed to make establishing links between devices more intuitive and user-friendly. With this technique, a target device can be identified by simply choosing the target device image on the user's terminal display. In this paper, we propose an efficient object recognition algorithm for point and-link communications for mobile terminals. Considering the limited capabilities of mobile terminals, the algorithm should operate in real time with low complexity and low memory requirements. Due to the bandwidth limit, the feature size of each device must be lower than 300 bytes. The proposed algorithm was optimized for each type of target including faces, devices and signboards.