#### DMCA

## A comparison and evaluation of multi-view stereo reconstruction algorithms. (2006)

### Cached

### Download Links

Venue: | In Proc. Computer Vision and Pattern Recognition ’06, |

Citations: | 529 - 14 self |

### Citations

1545 | A taxonomy and evaluation of dense two-frame stereo correspondence algorithms
- Scharstein, Szeliski
(Show Context)
Citation Context ...il recently similar. Here, however, a database of images with ground-truth results has made the comparison of algorithms possible and hence stimulated an even faster increase in algorithm performance =-=[1]-=-. In this paper, we aim to rectify this imbalance by providing, for the rst time, a collection of high-quality calibrated multi-view stereo images registered with groundtruth 3D models and an evaluat... |

1018 | A volumetric method for building complex models from range images
- Curless, Levoy
- 1996
(Show Context)
Citation Context ...ace characteristics and the viewing angle. For each object, roughly 200 individual scans were captured, aligned and merged on a 0:25mm grid, with the resulting mesh extracted with sub-voxel precision =-=[62]-=-; the accuracy of the combined scans is appreciably 0-7695-2646-2/06 $20.00 (c) 2006 IEEE greater than the individual scans. The procedure also produces per-vertex condence information, which we use ... |

566 | A theory of shape by space carving.
- Kutulakos, Seitz
- 1998
(Show Context)
Citation Context ...The vast majority of these measures operate by comparing pixels in one image to pixels in other images to see how well they correlate. For this reason, they are often called photo-consistencymeasures =-=[11]-=-. The choice of measure is not necessarily intrinsic to a particular algorithmit is often possible to take a measure from one method and substitute it in another. We categorize photo-consistency meas... |

491 |
The Visual Hull Concept for Silhouette-Based Image Understanding
- Laurentini
- 1994
(Show Context)
Citation Context ... forms of geometric reasoning to avoid oblique views and to minimize computations [5,11,26]. Another common quasi-geometric technique is to use a rough estimate of the surface such as the visual hull =-=[46]-=- to guess visibility for neighboring points [19, 47, 48]. The third type of method is to avoid explicit geometric reasoning and instead treat occlusions as outliers [31, 34, 37, 38]. Especially in cas... |

467 | Photorealistic scene reconstruction by voxel coloring.
- Seitz, Dyer
- 1997
(Show Context)
Citation Context ...projecting it into the input images, and evaluating the amount of mutual agreement between those projections. A simple measure of agreement is the variance of the projected pixels in the input images =-=[8, 11]-=-. Other methods compare images two at a time, and use windowmatching metrics such as sum of squared differences or normalized cross correlation [20, 23, 31]. An interesting feature of scene-space wind... |

379 | High-quality video view interpolation using a layered representation. - ZITNICK, KANG, et al. - 2004 |

316 | Multi-camera scene reconstruction via graph cuts
- Kolmogorov, Zabih
- 2002
(Show Context)
Citation Context ...a viewpoints. If the scene lies outside the convex hull of the camera centers, the occlusion ordering of points in the scene is same for all cameras [8], enabling a number of more efcient algorithms =-=[8, 10, 13, 35, 44]-=-. Quasi-geometric techniques use approximate geometric reasoning to infer visibility relationships. For example, a popular heuristic for minimizing the effects of occlusions is to limit the photo-cons... |

307 |
A multiple-baseline stereo”.
- Okutomi, Kanade
- 1993
(Show Context)
Citation Context ...e scene visibility can change dramatically with viewpoint, almost all modern multi-view stereo algorithms account for occlusions in some way or another. Early algorithms that did not model visibility =-=[6,27,43]-=- have trouble scaling to large distributions of viewpoints. Techniques for handling visibility include geometric, quasi-geometric, and outlier-based approaches. Geometric techniques seek to explicitly... |

280 |
Camera Calibration Toolbox for Matlab, http://www.vision.caltech.edu/bouguetj/ calib_doc,
- Bouguet
- 2005
(Show Context)
Citation Context ...ce of the object (the temple object is 10cm 16cm 8cm, and the dino is 7cm 9cm 7cm). The system was calibrated by imaging a planar calibration grid from 68 viewpoints over the hemisphere and using =-=[61]-=- to compute intrinsic and extrinsic parameters. From these parameters, we computed the camera's translational and rotational offset relative to the tip of the gantry arm, enabling us to determine the ... |

258 | A Maximum-Flow Formulation of the N-Camera Stereo Correspondence Problem,”
- Roy, Cox
- 1998
(Show Context)
Citation Context ...ximately view-independent intensities, i.e., Lambertian scenes, a number of new photo-consistency metrics have been devised that seek to model more general reflection functions (BRDFs) [15–17, 22, 23, 32]. Some methods also utilize silhouettes [27, 30, 31] or shadows [17, 42]. 2.3. Visibility model Visibility models specify which views to consider when evaluating photo-consistency measures. Because scene visibility can change dramatically with viewpoint, almost all modern multi-view stereo algorithms account for occlusions in some way or another. Early algorithms that did not model visibility [6,27,43] have trouble scaling to large distributions of viewpoints. Techniques for handling visibility include geometric, quasi-geometric, and outlier-based approaches. Geometric techniques seek to explicitly model the image formation process and the shape of the scene to determine which scene structures are visible in which images. A common approach in surface evolution approaches is to use the current estimate of the geometry to predict visibility for every point on that surface [5, 11, 12, 19, 20, 29, 30, 40]. Furthermore, if the surface evolution begins with a surface that encloses the scene volum... |

225 | Variational principles, surface evolution, pde’s, level set methods and the stereo problem
- Faugeras, Keriven
- 1998
(Show Context)
Citation Context ...ance of the projected pixels in the input images [8, 11]. Other methods compare images two at a time, and use windowmatching metrics such as sum of squared differences or normalized cross correlation =-=[20, 23, 31]-=-. An interesting feature of scene-space window-based methods is that the current estimate of the geometry can inform the size and shape of the window [20]. A number of other photo-consistency measures... |

189 | Multi-view stereo via volumetric graph-cuts.
- VOGIATZIS, TORR, et al.
- 2005
(Show Context)
Citation Context ... due to shadows. The 47 cameras corresponding to the ring dataset are shown in blue and red, and the 16 sparse ring cameras only in red. hull [46] that serves as an initial estimate of scene geometry =-=[5, 19, 31, 47, 48]-=-. Image-space algorithms [33, 3537] typically enforce constraints on the allowable range of disparity or depth values, thereby constraining scene geometry to lie within a near and far depth plane for... |

166 | Surface light fields for 3D photography,”
- Wood, Azuma, et al.
- 2000
(Show Context)
Citation Context ... ring cameras only in red. hull [46] that serves as an initial estimate of scene geometry [5, 19, 31, 47, 48]. Image-space algorithms [33, 35–37] typically enforce constraints on the allowable range of disparity or depth values, thereby constraining scene geometry to lie within a near and far depth plane for each camera viewpoint. 3. Multi-view data sets To enable a quantitative evaluation of multi-view stereo reconstruction algorithms, we collected several calibrated multi-view image sets and corresponding ground truth 3D mesh models. Similar data are available for surface lightfield studies [59, 60]; we have followed similar procedures for acquiring the images and models and for registering them to one another (although we add a step to automatically refine the alignment of the ground truth to the image sets based on minimizing photo-consistency). The surface lightfield data sets themselves are not, however, suitable for this evaluation due to the highly specular nature of the objects selected for those studies. We note that a number of other high quality multi-view datasets are publicly available (without registered ground truth models), and we provide links to many of these through our... |

164 | Robust parameter estimation in computer vision,”
- Stewart
- 1999
(Show Context)
Citation Context ...ion and Pattern Recognition (CVPR’06) 0-7695-2597-0/06 $20.00 © 2006 IEEE nation with other forms of geometric reasoning to avoid oblique views and to minimize computations [5,11,26]. Another common quasi-geometric technique is to use a rough estimate of the surface such as the visual hull [46] to guess visibility for neighboring points [19, 47, 48]. The third type of method is to avoid explicit geometric reasoning and instead treat occlusions as outliers [31, 34, 37, 38]. Especially in cases where scene points are visible more often than they are occluded, simple outlier rejection techniques [49] can be used to select the good views. A heuristic often used in tandem with outlier rejection is to avoid comparing views that are far apart, thereby increasing the likely percentage of inliers [31, 34, 37, 38]. 2.4. Shape prior Photo-consistency measures alone are not always sufficient to recover precise geometry, particularly in lowtextured scene regions [11, 50]. It can therefore be helpful to impose shape priors that bias the reconstruction to have desired characteristics. While priors are essential for binocular stereo, they play a less important role in multi-view stereo where the const... |

144 |
Constructing Virtual Worlds Using Dense Stereo.”
- Narayanan, Rander, et al.
- 1998
(Show Context)
Citation Context ...ometric reasoning to infer visibility relationships. For example, a popular heuristic for minimizing the effects of occlusions is to limit the photo-consistency analysis to clusters of nearby cameras =-=[31, 45]-=-. This approach is often used in combi0-7695-2646-2/06 $20.00 (c) 2006 IEEE nation with other forms of geometric reasoning to avoid oblique views and to minimize computations [5,11,26]. Another common... |

140 | Handling occlusions in dense multi-view stereo.
- Kang, Szeliski, et al.
- 2001
(Show Context)
Citation Context ...urface such as the visual hull [46] to guess visibility for neighboring points [19, 47, 48]. The third type of method is to avoid explicit geometric reasoning and instead treat occlusions as outliers =-=[31, 34, 37, 38]-=-. Especially in cases where scene points are visible more often than they are occluded, simple outlier rejection techniques [49] can be used to select the good views. A heuristic often used in tandem ... |

131 | Object-Centred Surface Reconstruction: Combining Multi-Image Stereo and Shading,”
- Fua, Leclerc
- 1995
(Show Context)
Citation Context ..., Lambertian scenes, a number of new photo-consistency metrics have been devised that seek to model more general reection functions (BRDFs) [1517, 22, 23, 32]. Some methods also utilize silhouettes =-=[27, 30, 31]-=- or shadows [17, 42]. 2.3. Visibility model Visibility models specify which views to consider when evaluating photo-consistency measures. Because scene visibility can change dramatically with viewpoin... |

97 | Stereo matching with transparency and matting. - Szeliski, Golland - 1999 |

94 | Light field mapping: Efficient representation and hardware rendering of surface light fields.
- Chen, Bouguet, et al.
- 2002
(Show Context)
Citation Context ... ring cameras only in red. hull [46] that serves as an initial estimate of scene geometry [5, 19, 31, 47, 48]. Image-space algorithms [33, 35–37] typically enforce constraints on the allowable range of disparity or depth values, thereby constraining scene geometry to lie within a near and far depth plane for each camera viewpoint. 3. Multi-view data sets To enable a quantitative evaluation of multi-view stereo reconstruction algorithms, we collected several calibrated multi-view image sets and corresponding ground truth 3D mesh models. Similar data are available for surface lightfield studies [59, 60]; we have followed similar procedures for acquiring the images and models and for registering them to one another (although we add a step to automatically refine the alignment of the ground truth to the image sets based on minimizing photo-consistency). The surface lightfield data sets themselves are not, however, suitable for this evaluation due to the highly specular nature of the objects selected for those studies. We note that a number of other high quality multi-view datasets are publicly available (without registered ground truth models), and we provide links to many of these through our... |

83 | Multi-View Stereo Revisited,”
- Goesele, Curless, et al.
- 2006
(Show Context)
Citation Context ... recover the 3D coordinates of salient feature points, then shrink a visual hull model so that the recovered points lie on its surface, then rene the result using energy minimization. Goesele et al. =-=[63]-=- compute a depth map from each camera viewpoint (similar to [31]) and merge the results using VRIP [62]. Hernandez and Schmitt [31] rst compute a depth map from each camera viewpoint and merge the re... |

79 | A Survey of Methods for Volumetric Scene Reconstruction from Photographs.
- Slabaugh, Culbertson, et al.
- 2001
(Show Context)
Citation Context ... and plan to extend our study to include such scenes in the future. This paper is not the rst to survey multi-view stereo algorithms; we refer readers to nice surveys by Dyer [3] and Slabaugh et al. =-=[4]-=- of algorithms up to 2001. However, the state of the art has changed dramatically in the last ve years, warranting a new overview of the eld. In addition, this paper provides the rst quantitative e... |

78 | Silhouette and Stereo Fusion for 3D Object Modeling,”
- Hernandez, Schmitt
- 2004
(Show Context)
Citation Context ...ance of the projected pixels in the input images [8, 11]. Other methods compare images two at a time, and use windowmatching metrics such as sum of squared differences or normalized cross correlation =-=[20, 23, 31]-=-. An interesting feature of scene-space window-based methods is that the current estimate of the geometry can inform the size and shape of the window [20]. A number of other photo-consistency measures... |

72 | Approximate N-view stereo.
- Kutulakos
- 2000
(Show Context)
Citation Context ...r of meshbased algorithms incorporate terms that cause triangles to shrink [29, 31] or prefer reference shapes such as a sphere or a plane [27]. Many methods based on voxel coloring and space carving =-=[5, 8, 9, 11, 12, 16, 18, 53]-=- instead prefer maximal surfaces. Since these methods operate by removing voxels only when they are not photo-consistent, they produce the largest photo-consistent scene reconstruction, known as the ... |

65 | Volumetric Scene Reconstruction from Multiple Views, in Foundations of Image Analysis,
- Dyer
- 2001
(Show Context)
Citation Context ...asets of specular scenes and plan to extend our study to include such scenes in the future. This paper is not the rst to survey multi-view stereo algorithms; we refer readers to nice surveys by Dyer =-=[3]-=- and Slabaugh et al. [4] of algorithms up to 2001. However, the state of the art has changed dramatically in the last ve years, warranting a new overview of the eld. In addition, this paper provides... |

64 | Multi-view reconstruction using photoconsistency and exact silhouette constraints: A maximum-flow formulation
- Sinha, Pollefeys
- 2005
(Show Context)
Citation Context ...iews and to minimize computations [5,11,26]. Another common quasi-geometric technique is to use a rough estimate of the surface such as the visual hull [46] to guess visibility for neighboring points =-=[19, 47, 48]-=-. The third type of method is to avoid explicit geometric reasoning and instead treat occlusions as outliers [31, 34, 37, 38]. Especially in cases where scene points are visible more often than they a... |

62 |
Modelling dynamic scenes by registering multi-view image sequences.
- Pons, Keriven, et al.
- 2005
(Show Context)
Citation Context ...an estimate of scene geometry to warp an image from one viewpoint to predict a different view. Comparing the predicted and measured images yields a photo-consistency measure known as prediction error =-=[26, 41]-=-. While prediction error is conceptually very similar to scene space measures, an important difference is the domain of integration. Scene space error functions are integrated over a surface and thus ... |

61 | Multi-view stereo beyond lambert.
- Jin, Soatto, et al.
- 2003
(Show Context)
Citation Context ...ance of the projected pixels in the input images [8, 11]. Other methods compare images two at a time, and use windowmatching metrics such as sum of squared differences or normalized cross correlation =-=[20, 23, 31]-=-. An interesting feature of scene-space window-based methods is that the current estimate of the geometry can inform the size and shape of the window [20]. A number of other photo-consistency measures... |

61 | Multi-view stereo reconstruction of dense shape and complex appearance. - Jin, Soatto, et al. - 2005 |

59 | Image-consistent surface triangulation. - Morris, Kanade - 2000 |

56 | Voxel carving for specular surfaces.
- Bonfort, Sturm
- 2003
(Show Context)
Citation Context ...r of meshbased algorithms incorporate terms that cause triangles to shrink [29, 31] or prefer reference shapes such as a sphere or a plane [27]. Many methods based on voxel coloring and space carving =-=[5, 8, 9, 11, 12, 16, 18, 53]-=- instead prefer maximal surfaces. Since these methods operate by removing voxels only when they are not photo-consistent, they produce the largest photo-consistent scene reconstruction, known as the ... |

53 | Shape reconstruction in projective grid space from large number of images.
- Saito, Kanade
- 1999
(Show Context)
Citation Context ...r of meshbased algorithms incorporate terms that cause triangles to shrink [29, 31] or prefer reference shapes such as a sphere or a plane [27]. Many methods based on voxel coloring and space carving =-=[5, 8, 9, 11, 12, 16, 18, 53]-=- instead prefer maximal surfaces. Since these methods operate by removing voxels only when they are not photo-consistent, they produce the largest photo-consistent scene reconstruction, known as the ... |

49 |
Prediction error as a quality metric for motion and stereo
- Szeliski
- 1999
(Show Context)
Citation Context ...an estimate of scene geometry to warp an image from one viewpoint to predict a different view. Comparing the predicted and measured images yields a photo-consistency measure known as prediction error =-=[26, 41]-=-. While prediction error is conceptually very similar to scene space measures, an important difference is the domain of integration. Scene space error functions are integrated over a surface and thus ... |

47 | Shape reconstruction from 3D and 2D data using PDE-based deformable surfaces. - Duan, Yang, et al. - 2004 |

47 | Bayesian 3D modeling from images using multiple depth maps.
- Gargallo, Sturm
- 2005
(Show Context)
Citation Context ...urface such as the visual hull [46] to guess visibility for neighboring points [19, 47, 48]. The third type of method is to avoid explicit geometric reasoning and instead treat occlusions as outliers =-=[31, 34, 37, 38]-=-. Especially in cases where scene points are visible more often than they are occluded, simple outlier rejection techniques [49] can be used to select the good views. A heuristic often used in tandem ... |

47 | A Bayesian method for probable surface reconstruction and decimation,
- Diebel, Thrun, et al.
- 2006
(Show Context)
Citation Context ...s is what enables many level-set algorithms to converge from a gross initial shape [20]. The preference for minimal surfaces can also result in a tendency to smooth over points of high curvature (see =-=[51, 52]-=- for ways to address this problem). Recent approaches based on volumetric min-cut [19, 47] also have a bias for minimum surfaces. A number of meshbased algorithms incorporate terms that cause triangle... |

46 | Multi-hypothesis, volumetric reconstruction of 3-D objects from multiple calibrated camera views.
- Eisert, Steinbach, et al.
- 1999
(Show Context)
Citation Context ...es works by iteratively evolving a surface to decrease or minimize a cost function. This class includes methods based on voxels, level sets, and surface meshes. Space carving [5, 11] and its variants =-=[9, 11, 12, 14, 18, 40, 53]-=- progressively remove inconsistent voxels from an initial volume. Other variants of this approach enable adding as well as deleting voxels to minimize an energy function [15, 54]. Level-set techniques... |

46 | Dealing with textureless regions and specular highlights - a progressive space carving scheme using a novel photo-consistency measure.
- Yang, Pollefeys, et al.
- 2003
(Show Context)
Citation Context ...ts [9, 11, 12, 14, 18, 40, 53] progressively remove inconsistent voxels from an initial volume. Other variants of this approach enable adding as well as deleting voxels to minimize an energy function =-=[15, 54]-=-. Level-set techniques minimize a set of partial differential equations dened on a volume. Like space carving methods, level-set methods typically start from a large initial volume and shrink inward;... |

44 | Methods for volumetric reconstruction of visual scenes.
- SLABAUGH, CULBERTSON, et al.
- 2004
(Show Context)
Citation Context ...rent estimate of the geometry can inform the size and shape of the window [20]. A number of other photo-consistency measures have been proposed to provide robustness to small shifts and other effects =-=[12, 18]-=-. Image space methods use an estimate of scene geometry to warp an image from one viewpoint to predict a different view. Comparing the predicted and measured images yields a photo-consistency measure ... |

44 | Tales of shape and radiance in multiview stereo.
- Soatto, Yezzi, et al.
- 2003
(Show Context)
Citation Context ... algorithmit is often possible to take a measure from one method and substitute it in another. We categorize photo-consistency measures based on whether they are dened in scene space or image space =-=[22]-=-. Scene space measures work by taking a point, patch, or volume of geometry, projecting it into the input images, and evaluating the amount of mutual agreement between those projections. A simple meas... |

42 | A multi-view approach to motion and stereo, in: - Szeliski - 1999 |

38 | Stochastic refinement of the visual hull to satisfy photometric and silhouette consistency constraints
- Isidoro, Sclaroff
- 2003
(Show Context)
Citation Context ...ntegration. Scene space error functions are integrated over a surface and thus often tend to prefer smaller surfaces, whereas prediction error is integrated over the set of images of a scene and thus ascribe more weight to parts of the scene that appear frequently or occupy a large image area. While most stereo algorithms have traditionally assumed approximately view-independent intensities, i.e., Lambertian scenes, a number of new photo-consistency metrics have been devised that seek to model more general reflection functions (BRDFs) [15–17, 22, 23, 32]. Some methods also utilize silhouettes [27, 30, 31] or shadows [17, 42]. 2.3. Visibility model Visibility models specify which views to consider when evaluating photo-consistency measures. Because scene visibility can change dramatically with viewpoint, almost all modern multi-view stereo algorithms account for occlusions in some way or another. Early algorithms that did not model visibility [6,27,43] have trouble scaling to large distributions of viewpoints. Techniques for handling visibility include geometric, quasi-geometric, and outlier-based approaches. Geometric techniques seek to explicitly model the image formation process and the shap... |

37 | Representing stereo data with the Delaunay triangulation. - Faugeras, Bras-Mehlman, et al. - 1990 |

35 | Example-Based Stereo with General BRDFs.
- Treuille, Hertzmann, et al.
- 2004
(Show Context)
Citation Context ...umber of new photo-consistency metrics have been devised that seek to model more general reection functions (BRDFs) [1517, 22, 23, 32]. Some methods also utilize silhouettes [27, 30, 31] or shadows =-=[17, 42]-=-. 2.3. Visibility model Visibility models specify which views to consider when evaluating photo-consistency measures. Because scene visibility can change dramatically with viewpoint, almost all modern... |

32 | Towards real-time voxel coloring.
- Prock, Dyer
- 1998
(Show Context)
Citation Context ...a viewpoints. If the scene lies outside the convex hull of the camera centers, the occlusion ordering of points in the scene is same for all cameras [8], enabling a number of more efcient algorithms =-=[8, 10, 13, 35, 44]-=-. Quasi-geometric techniques use approximate geometric reasoning to infer visibility relationships. For example, a popular heuristic for minimizing the effects of occlusions is to limit the photo-cons... |

30 | A probabilistic theory of occupancy and emptiness.
- Bhotika, Fleet, et al.
- 2002
(Show Context)
Citation Context ...es works by iteratively evolving a surface to decrease or minimize a cost function. This class includes methods based on voxels, level sets, and surface meshes. Space carving [5, 11] and its variants =-=[9, 11, 12, 14, 18, 40, 53]-=- progressively remove inconsistent voxels from an initial volume. Other variants of this approach enable adding as well as deleting voxels to minimize an energy function [15, 54]. Level-set techniques... |

25 | Surface Reconstruction from Feature Based Stereo. - Taylor - 2003 |

22 | Shape and view independent reflectance map from multiple views. - Yu, Xu, et al. - 2004 |

22 | Shadow carving.
- Savarese, Rushmeier, et al.
- 2001
(Show Context)
Citation Context ...umber of new photo-consistency metrics have been devised that seek to model more general reection functions (BRDFs) [1517, 22, 23, 32]. Some methods also utilize silhouettes [27, 30, 31] or shadows =-=[17, 42]-=-. 2.3. Visibility model Visibility models specify which views to consider when evaluating photo-consistency measures. Because scene visibility can change dramatically with viewpoint, almost all modern... |

20 | Variational stereovision and 3d scene flow estimation with statistical similarity measures. - Pons, Keriven, et al. - 2003 |

19 |
A probabilistic framework for the space carving algorithm.
- Broadhurst, Drummond, et al.
- 2001
(Show Context)
Citation Context ...a viewpoints. If the scene lies outside the convex hull of the camera centers, the occlusion ordering of points in the scene is same for all cameras [8], enabling a number of more efcient algorithms =-=[8, 10, 13, 35, 44]-=-. Quasi-geometric techniques use approximate geometric reasoning to infer visibility relationships. For example, a popular heuristic for minimizing the effects of occlusions is to limit the photo-cons... |

19 | Reconstructing relief surfaces.
- Vogiatzis, Torr, et al.
- 2004
(Show Context)
Citation Context ...g the geometry on a 3D domain, and the 2D representation is convenient particularly for smaller datasets. An alternative is to dene the depth maps relative to scene surfaces to form a relief surface =-=[39, 40]-=-. 2.2. Photoconsistency measure Numerous measures have been proposed for evaluating the visual compatibility of a reconstruction with a set of input images. The vast majority of these measures operat... |

19 | Reconstruction of scene models from sparse 3D structure. - Manessis, Hilton, et al. - 2000 |

17 | When is the shape of a scene unique given its light-field: A fundamental theorem of 3D vision?
- Baker, Sim, et al.
- 2003
(Show Context)
Citation Context ...ype of method is to avoid explicit geometric reasoning and instead treat occlusions as outliers [31, 34, 37, 38]. Especially in cases where scene points are visible more often than they are occluded, simple outlier rejection techniques [49] can be used to select the good views. A heuristic often used in tandem with outlier rejection is to avoid comparing views that are far apart, thereby increasing the likely percentage of inliers [31, 34, 37, 38]. 2.4. Shape prior Photo-consistency measures alone are not always sufficient to recover precise geometry, particularly in lowtextured scene regions [11, 50]. It can therefore be helpful to impose shape priors that bias the reconstruction to have desired characteristics. While priors are essential for binocular stereo, they play a less important role in multi-view stereo where the constraints from many views are stronger. Techniques that minimize scene-based photoconsistency measures naturally seek minimal surfaces with small overall surface area. This bias is what enables many level-set algorithms to converge from a gross initial shape [20]. The preference for minimal surfaces can also result in a tendency to smooth over points of high curvature ... |

14 |
Poxels: Probabilistic voxelized volume reconstruction
- DeBonnet, Viola
(Show Context)
Citation Context |

13 |
Image-based multiresolution shape recovery by surface deformation.
- Zhang, Seitz
- 2001
(Show Context)
Citation Context ...ch scene structures are visible in which images. A common approach in surface evolution approaches is to use the current estimate of the geometry to predict visibility for every point on that surface =-=[5, 11, 12, 19, 20, 29, 30, 40]-=-. Furthermore, if the surface evolution begins with a surface that encloses the scene volume and evolves by carving away that volume, this visibility approach can be shown to be conservative [11, 18];... |

13 | Geo-consistency for wide multi-camera stereo.
- Drouin, Trudeau, et al.
- 2005
(Show Context)
Citation Context ...urface such as the visual hull [46] to guess visibility for neighboring points [19, 47, 48]. The third type of method is to avoid explicit geometric reasoning and instead treat occlusions as outliers =-=[31, 34, 37, 38]-=-. Especially in cases where scene points are visible more often than they are occluded, simple outlier rejection techniques [49] can be used to select the good views. A heuristic often used in tandem ... |

13 | Improved voxel coloring via volumetric optimization. TR 3, Center for Signal and Image Processing,
- Slabaugh, Malzbender, et al.
- 2000
(Show Context)
Citation Context ...ts [9, 11, 12, 14, 18, 40, 53] progressively remove inconsistent voxels from an initial volume. Other variants of this approach enable adding as well as deleting voxels to minimize an energy function =-=[15, 54]-=-. Level-set techniques minimize a set of partial differential equations dened on a volume. Like space carving methods, level-set methods typically start from a large initial volume and shrink inward;... |

12 | Three-dimensional object reconstruction from two-dimensional images. - Rockwood, Winget - 1997 |

11 | Shape from multiple cues: Integrating local brightness information.
- Fromherz, Bichsel
- 1995
(Show Context)
Citation Context ...ch scene structures are visible in which images. A common approach in surface evolution approaches is to use the current estimate of the geometry to predict visibility for every point on that surface =-=[5, 11, 12, 19, 20, 29, 30, 40]-=-. Furthermore, if the surface evolution begins with a surface that encloses the scene volume and evolves by carving away that volume, this visibility approach can be shown to be conservative [11, 18];... |

11 | Progressive surface reconstruction from images using a local prior. In
- Zeng, Paris, et al.
- 2005
(Show Context)
Citation Context ...g the geometry on a 3D domain, and the 2D representation is convenient particularly for smaller datasets. An alternative is to dene the depth maps relative to scene surfaces to form a relief surface =-=[39, 40]-=-. 2.2. Photoconsistency measure Numerous measures have been proposed for evaluating the visual compatibility of a reconstruction with a set of input images. The vast majority of these measures operat... |

6 | A maximum-ow formulation of the n-camera stereo correspondence problem - Roy, Cox - 1998 |

6 | Poxels: Probabilistic voxelized volume reconstruction.
- Bonet, Viola
- 1999
(Show Context)
Citation Context ...ins with a surface that encloses the scene volume and evolves by carving away that volume, this visibility approach can be shown to be conservative [11, 18]; i.e., the set of cameras for which a scene point is predicted to be visible is a subset of the set of cameras in which that point is truly visible. Visibility computations can be simplified by constraining the allowable distribution of camera viewpoints. If the scene lies outside the convex hull of the camera centers, the occlusion ordering of points in the scene is same for all cameras [8], enabling a number of more efficient algorithms [8, 10, 13, 35, 44]. Quasi-geometric techniques use approximate geometric reasoning to infer visibility relationships. For example, a popular heuristic for minimizing the effects of occlusions is to limit the photo-consistency analysis to clusters of nearby cameras [31, 45]. This approach is often used in combiProceedings of the 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR’06) 0-7695-2597-0/06 $20.00 © 2006 IEEE nation with other forms of geometric reasoning to avoid oblique views and to minimize computations [5,11,26]. Another common quasi-geometric technique is to use ... |

5 | High-fidelity image-based modeling.
- Furukawa, Ponce
- 2006
(Show Context)
Citation Context ...iews and to minimize computations [5,11,26]. Another common quasi-geometric technique is to use a rough estimate of the surface such as the visual hull [46] to guess visibility for neighboring points =-=[19, 47, 48]-=-. The third type of method is to avoid explicit geometric reasoning and instead treat occlusions as outliers [31, 34, 37, 38]. Especially in cases where scene points are visible more often than they a... |

5 | Higher-order nonlinear priors for surface reconstruction.
- Tasdizen, Whitaker
- 2004
(Show Context)
Citation Context ...s is what enables many level-set algorithms to converge from a gross initial shape [20]. The preference for minimal surfaces can also result in a tendency to smooth over points of high curvature (see =-=[51, 52]-=- for ways to address this problem). Recent approaches based on volumetric min-cut [19, 47] also have a bias for minimum surfaces. A number of meshbased algorithms incorporate terms that cause triangle... |

4 |
Surface light ¯elds for 3D photography
- Wood, Azuma, et al.
- 2000
(Show Context)
Citation Context ...ti-view stereo reconstruction algorithms, we collected several calibrated multi-view image sets and corresponding ground truth 3D mesh models. Similar data are available for surface lighteld studies =-=[59, 60]-=-; we have followed similar procedures for acquiring the images and models and for registering them to one another (although we add a step to automatically rene the alignment of the ground truth to th... |

2 |
Stochastic renement of the visual hull to satisfy photometric and silhouette consistency constraints
- Isidoro, Sclaroff
(Show Context)
Citation Context ..., Lambertian scenes, a number of new photo-consistency metrics have been devised that seek to model more general reection functions (BRDFs) [1517, 22, 23, 32]. Some methods also utilize silhouettes =-=[27, 30, 31]-=- or shadows [17, 42]. 2.3. Visibility model Visibility models specify which views to consider when evaluating photo-consistency measures. Because scene visibility can change dramatically with viewpoin... |

1 |
et al. Multi-view stereo evaluation web
- Seitz
(Show Context)
Citation Context ...ion 5). While the current evaluation only includes methods whose authors were able to provide us their results by CVPR nal submission time, our datasets and evaluation results are publicly available =-=[2]-=- and open to the general community. We plan to regularly update the results, and publish a more comprehensive comparative evaluation as a full-length journal publication. We limit the scope of this pa... |

1 | Variational stereovision and 3D scene ow estimation with statistical similarity measures - Pons, Keriven, et al. - 2003 |

1 | Shape and view independent reectance map from multiple views - Yu, Xu, et al. - 2004 |

1 |
When is the shape of a scene unique given its light-eld: A fundamental theorem of 3D vision
- Baker, Sim, et al.
(Show Context)
Citation Context ...ng the likely percentage of inliers [31, 34, 37, 38]. 2.4. Shape prior Photo-consistency measures alone are not always sufcient to recover precise geometry, particularly in lowtextured scene regions =-=[11, 50]-=-. It can therefore be helpful to impose shape priors that bias the reconstruction to have desired characteristics. While priors are essential for binocular stereo, they play a less important role in m... |

1 |
Light eld mapping: Efcient representation and hardware rendering of surface light elds
- Chen, Bouguet, et al.
(Show Context)
Citation Context ...ti-view stereo reconstruction algorithms, we collected several calibrated multi-view image sets and corresponding ground truth 3D mesh models. Similar data are available for surface lighteld studies =-=[59, 60]-=-; we have followed similar procedures for acquiring the images and models and for registering them to one another (although we add a step to automatically rene the alignment of the ground truth to th... |

1 |
Multi-view stereo revisited. In CVPR,
- Goesele, Curless, et al.
- 2006
(Show Context)
Citation Context ...hm. In an effort to cover the current state of the art, we sought to include new, recently published algorithms rather than evaluating classic methods from a few years ago. In addition to the six reported here, three other groups tried out the data but were not able to produce reasonable results and are therefore not included in the study. Furukawa et al. [48] use wide-baseline stereo matching to recover the 3D coordinates of salient feature points, then shrink a visual hull model so that the recovered points lie on its surface, then refine the result using energy minimization. Goesele et al. [63] compute a depth map from each camera viewpoint (similar to [31]) and merge the results using VRIP [62]. Hernandez and Schmitt [31] first compute a depth map from each camera viewpoint and merge the results into a cost volume. They then iteratively deform a mesh, initialized at the visual hull, to find a minimum cost surface in this volume, also incorporating terms to fit silhouettes. Kolmogorov and Zabih [35] compute a set of depth maps using multi-baseline stereo with graph cuts, then merge the results into a voxel volume by computing the intersections of the occluded volumes from each viewp... |