#### DMCA

## Online 3d reconstruction using convex optimization (2011)

Venue: | in 1st Workshop on Live Dense Reconstruction From Moving Cameras (ICCV |

Citations: | 16 - 1 self |

### Citations

471 |
Parallel tracking and mapping for small AR workspaces”,
- Klein, Murray
- 2007
(Show Context)
Citation Context ...ive research area for years and has recently seen great progress. While in the pioneering work of [3] the authors use a statistical approach (Extended Kalman Filter), current state of the art systems =-=[4]-=- employ a geometric approach (Bundle Adjustment). In the latter work, the authors split the tracking and mapping tasks into distinct threads, thus making effective use of the nowubiquitous multi-core ... |

439 | Real-Time Simultaneous Localisation and Mapping with a Single Camera.
- Davison
- 2003
(Show Context)
Citation Context ... to construct the map. SLAM aims to solve these problems simultaneously. Realtime SLAM has been an active research area for years and has recently seen great progress. While in the pioneering work of =-=[3]-=- the authors use a statistical approach (Extended Kalman Filter), current state of the art systems [4] employ a geometric approach (Bundle Adjustment). In the latter work, the authors split the tracki... |

436 | A first-order primal-dual algorithm for convex problems with applications to imaging,
- Chambolle, Pock
- 2011
(Show Context)
Citation Context ...splitting the data and the regularization term and then performing an alternating minimization of easier subproblems. Here, we propose to use the recently introduced first-order primal-dual algorithm =-=[1]-=-. The advantage is that we can directly minimize (1) without performing any approximation. } (1)The primal-dual formulation of (1) is given by { ∫ N∑ ∫ } min max − u div p + λ h(x, i)|u(x) − di|dx u ... |

221 | A space-sweep approach to true multi-image matching,"
- Collins
- 1996
(Show Context)
Citation Context ...nferior for computing depthmaps. Whenever keyframes are added, depthmap generation is triggered in the reconstruction thread. We implemented a multiview version of the well known planesweep algorithm =-=[2]-=- using standard normalized cross correlation on the GPU. Additionally, we discard depth values with a correlation value below a threshold in order to get the most reliable depth hypotheses only. 2.2. ... |

107 | Live dense reconstruction with a single moving camera",
- Newcombe, Davison
- 2010
(Show Context)
Citation Context ...e feasible due to the computational complexity, although some interesting steps towards this direction have been made in [5]. The first dense, realtime-capable reconstruction approach is presented in =-=[7]-=-, where a base mesh is generated from the sparse point features of the SLAM system, which is successively refined via dense depth information obtained from realtime variational optical flow [9]. Since... |

88 | Improving the agility of keyframe-based SLAM.
- Klein, Murray
- 2008
(Show Context)
Citation Context ...ures. Obtaining a full 3D reconstruction by making the map dense seems not to be feasible due to the computational complexity, although some interesting steps towards this direction have been made in =-=[5]-=-. The first dense, realtime-capable reconstruction approach is presented in [7], where a base mesh is generated from the sparse point features of the SLAM system, which is successively refined via den... |

76 | H.: A Globally Optimal Algorithm for Robust TV-L1 Range Image Integration
- ZACH, POCK, et al.
(Show Context)
Citation Context ...limitations towards a full 3D-reconstruction as depicted in figure 1. We propose a system that uses state-of-the-art realtime tracking, GPU based depthmap generation and high quality depth map fusion =-=[11]-=- [10] in a realtime environment. Reconstruction is based on an energy minimization which can be efficiently computed on the GPU. As the camera explores the scene, data is added continuously to the rec... |

61 |
Anisotropic Huber-L1 optical flow.
- Werlberger, Trobin, et al.
- 2009
(Show Context)
Citation Context ...ted in [7], where a base mesh is generated from the sparse point features of the SLAM system, which is successively refined via dense depth information obtained from realtime variational optical flow =-=[9]-=-. Since the base mesh, once created, cannot be adapted arbitrarily (i.e. change topology), the method is inherently limited: concave objects (e.g. holes, overhangs etc.) are reconstructed only partial... |

41 | Fast and high quality fusion of depth maps.
- Zach
- 2008
(Show Context)
Citation Context ...ations towards a full 3D-reconstruction as depicted in figure 1. We propose a system that uses state-of-the-art realtime tracking, GPU based depthmap generation and high quality depth map fusion [11] =-=[10]-=- in a realtime environment. Reconstruction is based on an energy minimization which can be efficiently computed on the GPU. As the camera explores the scene, data is added continuously to the reconstr... |

27 |
Real-time dense geometry from a handheld camera
- Stuehmer, Gumhold, et al.
(Show Context)
Citation Context ..., once created, cannot be adapted arbitrarily (i.e. change topology), the method is inherently limited: concave objects (e.g. holes, overhangs etc.) are reconstructed only partially or not at all. In =-=[8]-=- the authors use multiple images from the PTAM framework to compute robust depth maps in realtime via variational optical flow. The resulting geometry is limited to 2.5D, since an integration/fusion o... |

16 |
A new median formula with applications to PDE based denoising
- Li, Osher
- 2009
(Show Context)
Citation Context ...of pointwise minimization problems. Given a number v(x), the solution of the proximal operator at a point x is defined as prox hist(v(x)) = arg min u { ‖u − v(x)‖ 2 2τ + λ N∑ } h(x, i)|u − di| i=1 In =-=[6]-=-, it is shown that the global minimizer of the above problem can be computed via a generalized shrinkage formula. It is given by prox hist(v) = median {d1, ..., dN , p0, ...., pN } where di are the di... |