## A Feature-Driven Approach to Locating Optimal Viewpoints for Volume Visualization (2005)

### Cached

### Download Links

- [visual.k.u-tokyo.ac.jp]
- [visual.k.u-tokyo.ac.jp]
- [nis-lab.is.s.u-tokyo.ac.jp]
- DBLP

### Other Repositories/Bibliography

Venue: | In IEEE Visualization |

Citations: | 30 - 1 self |

### BibTeX

@INPROCEEDINGS{Takahashi05afeature-driven,

author = {Shigeo Takahashi and Yuriko Takeshima},

title = {A Feature-Driven Approach to Locating Optimal Viewpoints for Volume Visualization},

booktitle = {In IEEE Visualization},

year = {2005},

pages = {495--502},

publisher = {IEEE Press}

}

### OpenURL

### Abstract

Figure 1: Locating optimal viewpoints by individually estimating the visibility quality of each feature subvolume. The value under each image represents its corresponding estimate normalized to [0.0, 1.0]. Optimal viewpoint selection is an important task because it considerably influences the amount of information contained in the 2D projected images of 3D objects, and thus dominates their first impressions from a psychological point of view. Although several methods have been proposed that calculate the optimal positions of viewpoints especially for 3D surface meshes, none has been done for solid objects such as volumes. This paper presents a new method of locating such optimal viewpoints when visualizing volumes using direct volume rendering. The major idea behind our method is to decompose an entire volume into a set of feature components, and then find a globally optimal viewpoint by finding a compromise between locally optimal viewpoints for the components. As the feature components, the method employs interval volumes and their combinations that characterize the topological transitions of isosurfaces according to the scalar field. Furthermore, opacity transfer functions are also utilized to assign different weights to the decomposed components so that users can emphasize features of specific interest in the volumes. Several examples of volume datasets together with their optimal positions of viewpoints are exhibited in order to demonstrate that the method can effectively guide naive users to find optimal projections of volumes.

### Citations

513 |
Smooth subdivision surfaces based on triangles
- Loop
- 1987
(Show Context)
Citation Context ...ference mesh (in red) that covers the viewing sphere (in blue) as shown in Figure 2(d). Actually, the reference mesh is constructed by subdividing an icosahedron twice using the Loop subdivision rule =-=[20]-=-, and thus offers a uniform distribution of 162 viewpoint samples on the sphere. The top and bottom color-coded disks in Figures 2(b) and (c) are the projections of the viewpoint entropy distributions... |

155 | R.: The contour spectrum
- BAJAJ, PASCUCCI, et al.
- 1997
(Show Context)
Citation Context ...values 2 . Here, our method takes advantage of an interface called an interval volume decomposer (IVD) [27], where each decomposed IV corresponds to a link of a level-set graph such as a contour tree =-=[3]-=-, which is obtained by tracking the topological transitions of isosurfaces with respect to the scalar field. The IVD constructs the level-set graph by detecting changes in the number of isosurface com... |

131 | Computing contour trees in all dimensions
- Carr, Snoeyink, et al.
(Show Context)
Citation Context ...is obtained by tracking the topological transitions of isosurfaces with respect to the scalar field. The IVD constructs the level-set graph by detecting changes in the number of isosurface components =-=[7]-=- and their associated topological types [22] according to the scalar field from an adaptive interpolation of voxel samples [28]. Note that the level-set graphs are usually too sensitive to noise espec... |

114 |
Mesh saliency
- Lee, Varshney, et al.
(Show Context)
Citation Context ...e probability distribution using the Shannon entropy. The viewpoint entropy works well and has been further applied to image-based rendering [33] and camera path planning [2]. Furthermore, Lee et al. =-=[18]-=- have recently developed a new saliency-based method of finding the viewpoint that maximizes the sum of the saliency for the visible faces of 3D meshes. However, the targets of these methods are still... |

89 |
The transfer function bake-off
- Pfister, Lorensen, et al.
- 2001
(Show Context)
Citation Context ...binations of IVs are employed as feature components. Here, Figures 9(a), (b), (c) and (d) show visualization results of the foot dataset [21], a laser fusion implosion dataset [25], the tooth dataset =-=[23]-=-, and the vortex dataset, respectively. Note that, for the cases in Figures 9(c) and (d), the same opacity TFs as in [6] are used to assign different weights to the decomposed IVs.s(a) (b) Figure 8: V... |

85 |
The singularities of the visual mappings
- Koenderink, Doorn
- 1976
(Show Context)
Citation Context ...n 6, Section 7 concludes this paper and refers to future work. 2 RELATED WORK One of the pioneering studies on viewpoint 1 evaluation is the aspect graph representation developed by Koenderink et al. =-=[17]-=-. This representation classifies the regions of the viewing sphere that surrounds the object into aspects, by identifying equivalent views of the object edges in a topological sense. The node of the a... |

67 |
A solution to the next best view problem for automated surface acquisition
- Pito
- 1999
(Show Context)
Citation Context ...her tackled by identifying the equivalent class of object silhouettes seen from available viewpoints [1], and by optimally distributing the view volumes of the visible surface to candidate viewpoints =-=[24]-=-. The configuration of such optimal 1 View directions for orthogonal projections are also referred to as viewpoints in this paper. viewpoints has been also investigated for acquiring a minimal set of ... |

65 | Failure to detect changes to attended objects in motion pictures
- Levin, Simons
- 1997
(Show Context)
Citation Context ...t of information embedded into the 2D projected image. The viewpoint selection is also important in that psychological studies show that the first view leaves a strong impression of the target volume =-=[19]-=-, although current GPU-based systems are available for real-time rotation and translation of the given volume. Actually, such methods of locating optimal viewpoints have been proposed especially for 3... |

62 | What object attributes determine canonical views
- Blanz, Tarr, et al.
- 1999
(Show Context)
Citation Context ... is to find the best arrangement of the feature components in an final projected image by avoiding their occlusions as many as possible. This is justified by the psychological studies of Blanz et al. =-=[5]-=-, where they describe the following three conditions for the viewpoint optimality: significance of visible features, stability of the view (with respect to small transformations), and the number of oc... |

56 |
3d object recognition using shape similarity-based aspect graph
- Cyr, Kimia
- 2001
(Show Context)
Citation Context ...presents an aspect of the target 3D object, and the link a visual transition between neighboring aspects. The aspect graph representation has been intensively studied as a tool for object recognition =-=[9]-=-, while its automatic computation is still the subject of ongoing research [26]. The aspect graph representation, however, can be applied only to rather simple polyhedra, and the associated resolution... |

52 | Automatic camera placement for image-based modeling
- Fleishman, Cohen-Or, et al.
- 2000
(Show Context)
Citation Context ...ections are also referred to as viewpoints in this paper. viewpoints has been also investigated for acquiring a minimal set of textures in the image synthesis techniques such as image-based rendering =-=[10, 15]-=-. However, these methods cannot be applied to our case directly because their objective is to recover the 3D surface shapes of target objects only. In the computer graphics applications, on the other ... |

46 |
A simple method for computing general position in displaying three-dimensional objects
- Kamada, Kawai
- 1988
(Show Context)
Citation Context ... of target objects only. In the computer graphics applications, on the other hand, several methods have been proposed to locate the best viewpoint for 3D polyhedral meshes. For example, Kamada et al. =-=[16]-=- defined an optimal viewpoint if it minimizes the number of degenerate faces under orthogonal projection, and Barrel et al. [4] extended this scheme in order to manage perspective projections effectiv... |

42 |
DE PANNE M.: Simplifying flexible isosurfaces with local geometric measures
- CARR, SNOEYINK, et al.
(Show Context)
Citation Context ...field from an adaptive interpolation of voxel samples [28]. Note that the level-set graphs are usually too sensitive to noise especially in acquired datasets and thus need some simplification process =-=[8, 28]-=- to delineate the global structure of the given volumes. Figure 4 shows a set of decomposed IVs of the proton hydrogenatom collision dataset together with the simplified level-set graph. As shown in t... |

40 | View selection for volume rendering
- Bordoloi, Shen
- 2005
(Show Context)
Citation Context ... of these methods are still limited to surfaces having zero thickness. The idea of locating optimal viewpoints for volumes has also been explored simultaneously but independently by Bordoloi and Shen =-=[6]-=-. Their method evaluates the balance between the contributions of voxels to pixels in the resultant image using the entropy function, and then finds stable viewpoints with respect to small transformat... |

37 |
W.: Automatic view selection using viewpoint entropy and its application to image-based modelling. Computer Graphics Forum 22
- Vázquez, Feixas, et al.
- 2004
(Show Context)
Citation Context ...d the optimal viewpoint by maximizing the information of the probability distribution using the Shannon entropy. The viewpoint entropy works well and has been further applied to image-based rendering =-=[33]-=- and camera path planning [2]. Furthermore, Lee et al. [18] have recently developed a new saliency-based method of finding the viewpoint that maximizes the sum of the saliency for the visible faces of... |

35 | TAKESHIMA Y.: Topological volume skeletonization and its application to transfer function design
- TAKAHASHI, FUJISHIRO
(Show Context)
Citation Context ...o assign different weights to the 32 extracted isosurfaces. Here, the opacity TF is designed so that it can emphasize the topological transitions of isosurfaces with respect to the scalar field. (See =-=[29, 34]-=-.) While the result seems to be better than the previous one, Figure 4: Level-set graph and its associated interval volume decomposition of the proton hydrogen-atom collision dataset. Four interval vo... |

34 |
Efficient computation of the topology of level sets
- Pascucci, Cole-McLaughlin
- 2002
(Show Context)
Citation Context ...nsitions of isosurfaces with respect to the scalar field. The IVD constructs the level-set graph by detecting changes in the number of isosurface components [7] and their associated topological types =-=[22]-=- according to the scalar field from an adaptive interpolation of voxel samples [28]. Note that the level-set graphs are usually too sensitive to noise especially in acquired datasets and thus need som... |

33 |
Maximum entropy light source placement
- Gumhold
- 2002
(Show Context)
Citation Context ...ing using the proposed criteria will be interesting when we quickly investigate inner structures of static and time-varying volume datasets. We can also incorporate entropy-based light planning phase =-=[13]-=- into the method so that we can integrate the arrangement of features with light source placement. Relationships of our formulation with human psychological preference in viewpoint selection should be... |

33 |
Planning for complete sensor coverage in inspection
- Tarbox, Gottschlich
- 1995
(Show Context)
Citation Context ...] found the combination of optimal viewpoints for object recognition by searching a tree structure that retains the collection of possible camera movements and their associated aspects. Tarbox et al. =-=[31]-=- developed a method of planning the optimal arrangement of sensors so that they completely surround the object surface for its precise shape reconstruction. The NBV problem has been further tackled by... |

28 |
Volumetric data exploration using interval volume
- Fujishiro, Maeda, et al.
- 1996
(Show Context)
Citation Context ...or evaluating a locally optimal viewpoint of each component so that we can find a global compromise between them. (See Figure 1.) As the feature components, the method employs an interval volume (IV) =-=[11, 14]-=-, which is a generalized concept of an isosurface and is defined as a subvolume obtained by sweeping an isosurface component within some range of scalar field values. Another advantage is that the met... |

24 |
Interval sets: a volume rendering technique generalizing isosurface extraction
- Guo
- 1995
(Show Context)
Citation Context ...or evaluating a locally optimal viewpoint of each component so that we can find a global compromise between them. (See Figure 1.) As the feature components, the method employs an interval volume (IV) =-=[11, 14]-=-, which is a generalized concept of an isosurface and is defined as a subvolume obtained by sweeping an isosurface component within some range of scalar field values. Another advantage is that the met... |

22 | Topological volume skeletonization using adaptive tetrahedralization
- Takahashi, Nielson, et al.
- 2004
(Show Context)
Citation Context ...vel-set graph by detecting changes in the number of isosurface components [7] and their associated topological types [22] according to the scalar field from an adaptive interpolation of voxel samples =-=[28]-=-. Note that the level-set graphs are usually too sensitive to noise especially in acquired datasets and thus need some simplification process [8, 28] to delineate the global structure of the given vol... |

21 |
Scene understanding techniques using a virtual camera
- Barral, Dorme, et al.
(Show Context)
Citation Context ...the best viewpoint for 3D polyhedral meshes. For example, Kamada et al. [16] defined an optimal viewpoint if it minimizes the number of degenerate faces under orthogonal projection, and Barrel et al. =-=[4]-=- extended this scheme in order to manage perspective projections effectively. While these methods offer good viewpoints, their use is still limited to polygonal meshes with a small number of faces. On... |

20 | Planning Multiple Observations for Object Recognition
- Gremban, Ikeuchi
- 1992
(Show Context)
Citation Context ...rrounds the whole surface of the target object. This problem of planning camera positions is referred to as the next best view (NBV) problem in computer vision literature. For example, Gremban et al. =-=[12]-=- found the combination of optimal viewpoints for object recognition by searching a tree structure that retains the collection of possible camera movements and their associated aspects. Tarbox et al. [... |

15 |
Way-finder: guided tours through complex walkthrough models
- Andújar, Vázquez, et al.
- 2004
(Show Context)
Citation Context ...imizing the information of the probability distribution using the Shannon entropy. The viewpoint entropy works well and has been further applied to image-based rendering [33] and camera path planning =-=[2]-=-. Furthermore, Lee et al. [18] have recently developed a new saliency-based method of finding the viewpoint that maximizes the sum of the saliency for the visible faces of 3D meshes. However, the targ... |

13 | Finite-resolution aspect graphs of polyhedral objects
- Shimshoni, Ponce
- 1997
(Show Context)
Citation Context ...etween neighboring aspects. The aspect graph representation has been intensively studied as a tool for object recognition [9], while its automatic computation is still the subject of ongoing research =-=[26]-=-. The aspect graph representation, however, can be applied only to rather simple polyhedra, and the associated resolution on the viewing sphere is not high enough to locate the optimal viewpoint preci... |

12 | 14.9 tflops three-dimensional fluid simulation for fusion science with hpf on the earth simulator
- Sakagami, Murai, et al.
(Show Context)
Citation Context ...n Figure 9 where the combinations of IVs are employed as feature components. Here, Figures 9(a), (b), (c) and (d) show visualization results of the foot dataset [21], a laser fusion implosion dataset =-=[25]-=-, the tooth dataset [23], and the vortex dataset, respectively. Note that, for the cases in Figures 9(c) and (d), the same opacity TFs as in [6] are used to assign different weights to the decomposed ... |

12 |
Interval Volume Decomposer: A Topological Approach to Volume Traversal
- Takahashi, Fujishiro, et al.
- 2005
(Show Context)
Citation Context ...ined by integrating a connected component of an isosurface within some range of the scalar field values 2 . Here, our method takes advantage of an interface called an interval volume decomposer (IVD) =-=[27]-=-, where each decomposed IV corresponds to a link of a level-set graph such as a contour tree [3], which is obtained by tracking the topological transitions of isosurfaces with respect to the scalar fi... |

10 | Introducing topological attributes for objective-based visualization of simulated datasets
- Takeshima, Takahashi, et al.
- 2005
(Show Context)
Citation Context ... with the viewpoint location has been enhanced by replacing a simple single-dimensional opacity TF (Figure 5(a)) with a multi-dimensional opacity TF that emphasizes the inner structures in the volume =-=[30]-=- (Figure 5(b)). 5.4 Formulation of the Viewpoint Entropy with the Combinations of Neighboring Interval Volumes As seen in the previous figures, the individual IVs serve as reasonable feature component... |

7 |
Automatic view selection in multi-view object recognition
- Abbasi, Mokhtarian
- 2000
(Show Context)
Citation Context ...ely surround the object surface for its precise shape reconstruction. The NBV problem has been further tackled by identifying the equivalent class of object silhouettes seen from available viewpoints =-=[1]-=-, and by optimally distributing the view volumes of the visible surface to candidate viewpoints [24]. The configuration of such optimal 1 View directions for orthogonal projections are also referred t... |

4 |
Viewpoint selection using view entropy
- Vazquez, Feixas, et al.
- 2001
(Show Context)
Citation Context ...viewpoints, their use is still limited to polygonal meshes with a small number of faces. One solution for the case of general 3D meshes is the viewpoint entropy, which is formulated by Vázquez et al. =-=[32]-=- to evaluate the balance of visible faces in 2D projected images. In this formulation, they consider the visibility of each face as its probability, and find the optimal viewpoint by maximizing the in... |

2 |
PC-based real-time texture painting on real world objects
- Iwakiri, Kaneko
- 2001
(Show Context)
Citation Context ...ections are also referred to as viewpoints in this paper. viewpoints has been also investigated for acquiring a minimal set of textures in the image synthesis techniques such as image-based rendering =-=[10, 15]-=-. However, these methods cannot be applied to our case directly because their objective is to recover the 3D surface shapes of target objects only. In the computer graphics applications, on the other ... |

2 |
Automating transfer function design based on topology analysis
- Weber, Scheuermann
- 2004
(Show Context)
Citation Context ...o assign different weights to the 32 extracted isosurfaces. Here, the opacity TF is designed so that it can emphasize the topological transitions of isosurfaces with respect to the scalar field. (See =-=[29, 34]-=-.) While the result seems to be better than the previous one, Figure 4: Level-set graph and its associated interval volume decomposition of the proton hydrogen-atom collision dataset. Four interval vo... |