Human-centered picture slideshow personalization for mobile devices

This paper presents a human-centered picture slideshow system for mobile users. In contrast to conventional ROIs (region-of-interest) detection based systems, we provide mobile users the freedom of personalizing ROIs in a convenient and effective way. Here, we import a simple human interaction, i.e., only a single click, to give a hint for users´ ROIs. First, local saliency map (LSM) is generated, which considers not only multi-scale contrast, but also the self-correlation measure and central effect. Then a local fuzzy growing method is adopted to extract ROIs automatically based on LSM. Extensive experiments and user studies show the encouraging performance of the proposed system.