We present a 3D reconstruction method to estimate depth maps using light-field camera. The estimated depth is widely used for photographic editing. The proposed algorithm, as a feature-based method, provides high quality depth map. First, the distinctive features are extracted from reference image, and the correspondences are tracked. Second, the depth values for each points are calculated by small angle approximated bundle adjustment. Finally, the whole depth map is estimated by simple depth propagation method. To show the effectiveness of our algorithm, we compare our results with the state-of-the-art algorithm.
|저 자||Sunghoon Im, Hae-Gon Jeon, Hyowon Ha, In So Kweon|
|학 회||The 12th International Conference on Ubiquitous Robots and Ambient Intelligence (URAI)|
|Notes||Acknowledgment [NCRC]; The first author was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIP) (No. 2010-0028680).|