Where Should I Stand? Learning Based Human Position Recommendation for Mobile Photographing

Pengfei Xu,Hongxun Yao,Rongrong Ji,Xian-Ming Liu,Xiaoshuai Sun

DOI: https://doi.org/10.1007/s11042-012-1343-2

IF: 2.577

2013-01-01

Multimedia Tools and Applications

Abstract:In this paper, we study the problem of human position recommendation in mobile photographing and propose a learning-based method to summarize the photographing knowledge from massive social images to improve the robustness and effectiveness. In contrast to existing photographing guide methods, we focus on turning to the collaborative web data source and learning the distribution of human position. To overcome the challenges in landmark image alignment and the relative human position projection, we propose a 3D reconstruction-based method to align the background region and human region into a uniform coordinate system. Finally, a camera-view sensitive human position recommendation strategy is carried out. A dataset containing 30,000 photos of ten landmark scenes is collected from Flickr, and a group of experiments are conducted comparing both our alternatives and various other baseline methods. Moreover, an application is developed on mobile phones to implement the real-time photographing recommendation. The experimental results show that our proposed framework achieves promising results, which demonstrate the robustness and effectiveness of our approach.