Depth estimation for light field camera using multi-cue integrated cost volume = 다중 정보 융합 비용 체적을 이용한 라이트 필드 카메라에서의 깊이 지도 추정

Cited 0 time in webofscience Cited 0 time in scopus
  • Hit : 389
  • Download : 0
With the recent introduction of commercial light field cameras, depth estimation using a light field camera draw significant interest of researchers. However, since the baseline of a light field camera is restricted to the size of lens aperture and its sub-aperture images are quite noisy, depth map estimation of a light field camera is a challenging problem. In this thesis, a robust depth map estimation method using multi-cue integrated cost volume is suggested. By using the cost volume concept, fast computation of depth map is achieved. To alleviate the ambiguity problem that comes from weakly textured region and narrow baseline of sub-aperture images, the pixels with reliable depth values are picked out and used to correct nearby unreliable depth values. A focus cue is exploited to filter out the pixels with unreliable depth values. A pixel with a depth value that falls within a confidence interval defined by the focus cue is classified as the anchor pixel having a reliable depth value. Using the estimated reliable depth values, unreliable depth values are corrected with an assumption that nearby pixels with similar color values may have similar depth values. Such an assumption significantly improves the performance of a non-convex optimization step. A discrete-continuous optimizer is adopted to obtain the final depth map by minimizing an objective function consisting of the color-gradient consistency and the depth difference between nearby pixels. The discrete-continuous optimization is suitable for efficiently combining the discrete cost volume structure and the continuous depth smoothness term at the same time. The performance of the proposed method is systematically validated on both indoor and outdoor datasets. The experimental results show that the proposed method outperforms other existing methods.
Advisors
Kweon, In-Soresearcher권인소
Description
한국과학기술원 : 전기및전자공학과,
Publisher
한국과학기술원
Issue Date
2014
Identifier
569218/325007  / 020123085
Language
eng
Description

학위논문(석사) - 한국과학기술원 : 전기및전자공학과, 2014.2, [ iii, 42 p. ]

Keywords

Light field camera; 이산-연속 최적화; 비용 체적; 깊이 지도; 라이트 필드 카메라; Discrete-continuous optimization; Depth map; Cost volume

URI
http://hdl.handle.net/10203/196746
Link
http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=569218&flag=dissertation
Appears in Collection
EE-Theses_Master(석사논문)
Files in This Item
There are no files associated with this item.

qr_code

  • mendeley

    citeulike


rss_1.0 rss_2.0 atom_1.0