#### DMCA

## Near real-time stereomatching using geodesic diffusion (2012)

Venue: | IEEE Transactions on Pattern Analysis andMachine Intelligence |

Citations: | 9 - 0 self |

### Citations

1881 | Scale-space and edge detection using anisotropic diffusion
- Perona, Malik
- 1990
(Show Context)
Citation Context ...esults produced by adaptive-weight algorithms, there is a major problem with their implementation. In fact, pixel-wise support-weight computation is a highly timeconsuming task. Anisotropic diffusion =-=[9]-=-, a computer vision technique very similar to adaptive weighting but computationally less expensive, has been used in an attempt to reduce the computational requirements. Nevertheless, the implementat... |

1545 | A Taxonomy and Evaluation of Dense Two-Frame Stereo Correspondence Algorithms
- Scharstein, Szeliski
- 2002
(Show Context)
Citation Context ...tching is one of the most active areas in the field of computer vision. As a consequence, a variety of approaches have been proposed. An excellent survey of stereo matching algorithms can be found in =-=[1]-=-. According to this survey, most stereo algorithms can be categorized into two major classes: local methods and global methods. Local approaches use information within a finite region around the pixel... |

1152 | Bilateral filtering for gray and color images
- Tomasi, Manduchi
- 1998
(Show Context)
Citation Context ...hing measures were tested with each aggregation technique: TAD and HMI. The input stereo images were filtered for all of the aggregation methods analyzed in this section using a 5 5 bilateral filter =-=[21]-=- for weight computation (not for DSI computation). The reason for this choice is that high-frequency information is a very useful cue for pixel matching. In contrast, for weight computation, it is mor... |

340 | A Stereo Matching Algorithm with an Adaptive Window: Theory and Experiment,
- Kanade, Okutomi
- 1991
(Show Context)
Citation Context ... best support window from a set of predefined windows [2], [3]. Variablewindow methods do not select the best window from a predefined set, but rather compute an optimal support window for each pixel =-=[4]-=-, [5], [6]. Adaptive-weight methods compute a support weight for each pixel inside a fixed-size square window [7], [8]. Adaptive-weight methods are the local algorithms yielding the best results, comp... |

218 | Stereo processing by semiglobal matching and mutual information,”
- Hirschmüller
- 2008
(Show Context)
Citation Context ...nsformation: wðp; qÞ exp Dðp; qÞ : ð6Þ The pixel-based matching cost measure used in [8] is hierarchical mutual information (HMI). The implementation of this matching measure is described in =-=[14]-=-, to which we refer the reader for details as this report is focused on aggregation techniques and a detailed description of HMI would unnecessarily lengthen our discussion. The two previously describ... |

156 | Adaptive support-weight approach for correspondence search
- Yoon, Kweon
- 2006
(Show Context)
Citation Context ...dow from a predefined set, but rather compute an optimal support window for each pixel [4], [5], [6]. Adaptive-weight methods compute a support weight for each pixel inside a fixed-size square window =-=[7]-=-, [8]. Adaptive-weight methods are the local algorithms yielding the best results, comparable to those generated using global optimization techniques. Here, the weights regulate each pixel’s influence... |

143 | Large occlusion stereo.
- Bobik, Intille
- 1999
(Show Context)
Citation Context ...niques have been proposed to choose the support for each pixel to satisfy this “same-disparity condition.” Multiple-window methods select the best support window from a set of predefined windows [2], =-=[3]-=-. Variablewindow methods do not select the best window from a predefined set, but rather compute an optimal support window for each pixel [4], [5], [6]. Adaptive-weight methods compute a support weigh... |

126 | Stereo matching with nonlinear diffusion.
- Scharstein, Szeliski
- 1998
(Show Context)
Citation Context ...e of small windows was compensated for by applying several iterations of the adaptive-weight aggregation step. Other diffusion techniques have been applied to stereo matching. Scharstein and Szeliski =-=[15]-=- tested various diffusion models. Most of the proposed methods rely on local diffusion (costs are diffused along constant-disparity planes). However, the results of these local diffusion solutions are... |

97 | Efficient stereo with multiple windowing,”
- Fusiello, Roberto, et al.
- 1997
(Show Context)
Citation Context ... techniques have been proposed to choose the support for each pixel to satisfy this “same-disparity condition.” Multiple-window methods select the best support window from a set of predefined windows =-=[2]-=-, [3]. Variablewindow methods do not select the best window from a predefined set, but rather compute an optimal support window for each pixel [4], [5], [6]. Adaptive-weight methods compute a support ... |

81 | Fast variable window for stereo correspondence using integral images,” in
- Veksler
- 2003
(Show Context)
Citation Context ...ort window from a set of predefined windows [2], [3]. Variablewindow methods do not select the best window from a predefined set, but rather compute an optimal support window for each pixel [4], [5], =-=[6]-=-. Adaptive-weight methods compute a support weight for each pixel inside a fixed-size square window [7], [8]. Adaptive-weight methods are the local algorithms yielding the best results, comparable to ... |

59 |
High-Quality Real-Time Stereo Using Adaptive Cost Aggregation and Dynamic
- Wang, Liao, et al.
- 2006
(Show Context)
Citation Context ...g algorithms with smaller computational complexities have been proposed that try to simulate the behavior of the adaptive-weight algorithm [7] and the geodesic supportweight approach [8]. Wang et al. =-=[16]-=- proposed a two-pass aggregation scheme to approximate adaptive weights and enable a GPU implementation. Richardt et al. [17] recently proposed a new real-time local stereo algorithm that uses a bilat... |

58 | A variable window approach to early vision.
- Boykov, Veksler, et al.
- 1998
(Show Context)
Citation Context ... support window from a set of predefined windows [2], [3]. Variablewindow methods do not select the best window from a predefined set, but rather compute an optimal support window for each pixel [4], =-=[5]-=-, [6]. Adaptive-weight methods compute a support weight for each pixel inside a fixed-size square window [7], [8]. Adaptive-weight methods are the local algorithms yielding the best results, comparabl... |

52 | A performance study on different cost aggregation approaches used in real-time stereo matching ”,
- Gong, W, et al.
- 2007
(Show Context)
Citation Context ...esults and conclusions in Sections 4 and 5, respectively. 2 RELATED WORK Our approach was inspired by recent advances in the area of variable cost aggregation supports that were recently evaluated in =-=[12]-=- and [13]. Among these solutions, the adaptive-weight algorithm proposed by Yoon and Kweon [7] deserves particular attention for its results. Hosni et al. [8] proposed a modification in the weight-com... |

40 | Clasification and evaluation of cost aggregation methods for stereo correspondence”. In
- Tombari, Mattoccia, et al.
- 2008
(Show Context)
Citation Context ...d conclusions in Sections 4 and 5, respectively. 2 RELATED WORK Our approach was inspired by recent advances in the area of variable cost aggregation supports that were recently evaluated in [12] and =-=[13]-=-. Among these solutions, the adaptive-weight algorithm proposed by Yoon and Kweon [7] deserves particular attention for its results. Hosni et al. [8] proposed a modification in the weight-computation ... |

35 | Real-time spatiotemporal stereo matching using the dualcross-bilateral grid
- Richardt, Orr, et al.
- 2010
(Show Context)
Citation Context ...ght algorithm [7] and the geodesic supportweight approach [8]. Wang et al. [16] proposed a two-pass aggregation scheme to approximate adaptive weights and enable a GPU implementation. Richardt et al. =-=[17]-=- recently proposed a new real-time local stereo algorithm that uses a bilateral grid for cost aggregation, achieving a 200 times improvement in speed over the original adaptive-weight algorithm. Essen... |

33 | Segmentation-based adaptive support for accurate stereo correspondence”.
- Tombari, Mattoccia, et al.
- 2007
(Show Context)
Citation Context ...daptive weights but that produce more accurate results, even if the execution time of these solutions is higher than the computation time of the original adaptive-weight algorithm [7]. Tombari et al. =-=[19]-=- proposed the use of segmentation along with adaptive weights. 3 GEODESIC DIFFUSION Our algorithm was directly inspired by anisotropic diffusion [9], an iterative computer vision technique in which th... |

29 | Local Stereo Matching Using Geodesic Support Weights,”
- Hosni, Bleyer, et al.
- 2009
(Show Context)
Citation Context ...wðp; qÞ transforms the cost of the minimum cost path into a weight with a parameter controlling this transformation: wðp; qÞ exp Dðp; qÞ : ð6Þ The pixel-based matching cost measure used in =-=[8]-=- is hierarchical mutual information (HMI). The implementation of this matching measure is described in [14], to which we refer the reader for details as this report is focused on aggregation technique... |

22 |
Cost Aggregation and Occlusion Handling with WLS in Stereo Matching,”
- Min, Sohn
- 2008
(Show Context)
Citation Context ...pensive, has been used in an attempt to reduce the computational requirements. Nevertheless, the implementation of this technique for stereo matching has so far resulted in low-quality disparity maps =-=[10]-=-. Min and Sohn [10] proposed an intermediate solution: They used adaptive-weight aggregation windows (smaller than those used in adaptive-weight (AW) solutions) and incorporated the iterative nature o... |

6 | Near real-time stereo with adaptive support weight approaches
- Hosni, Bleyer, et al.
(Show Context)
Citation Context ...algorithms [16], [17] are less accurate than those of the adaptiveweight algorithm [7]. This is also the case for a faster version of the geodesic support-weight approach [8] proposed by Hosni et al. =-=[18]-=-; the simplified version is much faster but has reduced accuracy. Despite this general complexity-reduction trend, some authors have proposed algorithms based on adaptive weights but that produce more... |

2 |
Support Aggregation via NonLinear Diffusion with Disparity-Dependent Support-Weights for Stereo Matching
- Yoon, Jeong, et al.
- 2009
(Show Context)
Citation Context ...eddy, and Cones. Our geodesic diffusion approach is compared with three other aggregationmethods: adaptive weights [7], geodesic adaptive weights (GAW) [8], and iterative adaptive weights (IAW) [10], =-=[11]-=- (these last two publications both proposed a similar implementation of iterative adaptive weights, with very similar results). For IAW, 5 5 and 9 9 windows were used in our tests as proposed, respe... |

1 |
Middlebury Stereo Evaluation—Version 2
- Scharstein, Szeliski
- 2011
(Show Context)
Citation Context ...doubled. Finally, new parameters should be included to model 45 or 135 turns in the diffusion direction. 4 EXPERIMENTAL RESULTS 4.1 Quality of the Results The Middlebury stereo evaluation website =-=[20]-=- provides a convenient framework for evaluating the accuracy of the reconstruction by the percentage of bad pixels in the computation of four disparity maps using four stereo pairs named Tsukuba, Venu... |