This paper aims to put forward a mobile augmented reality scheme by combining the C/S (client/server) architecture and BRF (boosted random ferns) algorithm that can enable the recognition performance for product packaging. BRF was an effective and robust feature matching algorithm, but not suitable for mobile phones directly because of the devices’ limited capabilities. This paper combined the C/S mode and BRF algorithm for matching features, and performed experiments and compared the recognition speed and accuracy of CS-BRF and ORB. Experimental results showed that CS-BRF had close efficiency and better accuracy than ORB. In conclusion, CS-BRF can recognize printed images rapidly and precisely, and thus is well applicable to mobile augmented reality system for product packaging.