Resource-limited intelligent photo management on mobile platforms

Abstract

Apart from wireless communication issues, a key technical challenge is how to achieve the best-experienced photo browsing given the limited screen size of the mobile devices. Therefore, in this paper, we propose a novel technique, resource — limited intelligent photo management (RIPM), on the demand of reducing the complexity of computation on Android mobile platform, in which photos captured are analyzed directly in JPEG compressed domain and are further classified in a real-time manner based on the human subject's gender. In order to make the system robust to luminance variations, DC coefficients are discarded. In addition, for the low-complexity purpose and the effective gender discrimination, a set of AC coefficients are selected automatically based on a three-step dimensionality reduction, in which evaluation of the coefficients' significance is conducted by LDA-based approach. Experimental results obtained by using extensive dataset captured under uncontrolled environments show that our system is effective for photo managements on resource-limited mobile platform.