| Stenstrom, J.R., "Constructing Object Models from Multiple Images", Intl. J. of Computer Vision, 9, p185-212, 1992. |
....Methods that use a mesh surface to model and integrate each of a set of range images, such as work by Turk and Levoy [23] or by Rutishauser et al. 16] or to model a complete point sampling as by Hoppe [8] or Fua and Sander [7] have also proven useful in this task. Both Connolly and Stenstrom [4,17] and Martin and Aggarwal [12] perform edge detection and projection from intensity images, a concept that is revisited by Laurentini [11] Curless and Levoy [5] present a system that uses a mesh in a ray casting operation to weight voxels in an octree, which is then used as input to an isosurface ....
J. Stenstrom, C.I. Connolly, Constructing object models from multiple images, Int. J. Comput. Vis. 9 (3) (1992) 185--212.
....image features can be used for resolving matches. However, the key process of linking up relevant features into self consistent models is usually ad hoc and unreliable. Recently, several algorithms capable of analyzing long baseline inputs, or images from distant viewpoints have been proposed [43], 8] 12] 18] 14] 38] In these systems, the camera pose information is assumed to be given. Nevertheless, the correspondence problem is made more challenging due to the potential for significant viewpoint changes. One feasible solution involves performing a space sweep search in the 3D ....
Stenstrom, J.R., "Constructing Object Models from Multiple Images", Intl. J. of Computer Vision, 9, p185-212, 1992. BIBLIOGRAPHY 157
....as in [3] In [21] a limited number of model primitives are hypothesized from the single view SAG. Single view SAGs may be augmented by occlusion surfaces to produce an OPUS (Object Plus Unseen Space) model [13] or a space envelope model [16] or enclosed using a hypothesized back projection [28]. SAGs may be combined from separate views to produce a complete SAG of an object, as in [7, 22] 3 In all these works, the convention has been to refer to the surface model as a boundary representation (b rep) The nature of the input data (i.e. the scenes imaged) supports modeling by ....
....supports modeling by relatively large surface patches, having planar or simple quadratic equations. The SAG paradigm assumes that some (or all) of the imaged scene may be modeled by unique descriptors (i.e. commonly recognized surface patches) Thus to a large degree, all the models produced in [3, 7, 13, 16, 21, 22, 28] are viewpoint independent (i.e. there is an obvious one to one correlation between scene surfaces and model surfaces, which would persist from neighboring viewpoints) B reps generally model the image on the order of less than 1 2 the size of the original data 3 . 2. In the mesh ....
J. R. Stenstrom and C. I. Connolly, "Constructing Object Models from Multiple Images ", in International Journal of Computer Vision, vol. 9, no. 3, 1992, pp. 185-212. 15
....multiframe stereo) A third approach is to use other information which is implicitly present in the profiles. The profile of a surface from a given view determines a bounding cone. The intersection of these cones from different viewpoints can be used to construct the bounding volume for an object [26, 27]. Merging of the appropriate pieces of these two surfaces along the frontier and the natural boundary can produce a closed surface which uses all of the information available from the profiles. Additional information from surface markings can potentially be combined for visible regions which are ....
J.R.Steenstrom and C.I.Connolly, `Constructing object models from multiple images', Int. J. of Computer Vision, 9 (1992), 185-212.
....stereo) A third approach is to use other information which is implicitly present in the profiles. The profile of a surface from a given view determines a bound ing cone. The intersection of these cones from different viewpoints can be used to construct the bounding volume for an object [12, 13]. Merging of the appropriate pieces of these two surfaces along the frontier and the natural boundary can produce a closed surface which uses all of the information available from the profiles. Additional information from surface markings can potentially be combined for visible regions which are ....
J.R.Steenstrom and C.I.Connolly, `Constructing object models from multiple images ', Int. J. of Computer Vision, 9 (1992), 185-212.
....1. Objects are not encountered free floating in isolation. Even when viewing an object under artificially isolating conditions, such as on a turn table or assembly line, the background surfaces (supporting surface, floor, walls, etc. must be manually or heuristically removed (see, for instance, [9, 44]) While such operations may be adequate for some applications, such as the reverse engineering of parts, they do not suffice for model construction in an unstructured environment. 2. The complete boundary of an object (or objects) is not visible in any single view. Various approaches have been ....
....The scene column indicates which modeling schemes have been demonstrated upon images containing unstructured groups of objects, that have not been artificially isolated from any background or from each other, such as the kitchen image in Figure 1. Previously proposed b rep construction schemes [13, 35, 44], including our own previous work on the OPUS b rep [22] are not applicable to this type of image. Previously proposed mesh construction schemes [6, 9, 11, 20, 43] have not been demonstrated upon this type of image. Closed quadric representations [15, 28, 33, 37, 38, 42, 45] including ....
[Article contains additional citation context not shown here]
J. R. Stenstrom and C. I. Connolly, "Constructing Object Models from Multiple Images ", in International Journal of Computer Vision, vol. 9, no. 3, 1992, pp. 185-212.
No context found.
Stenstrom, J.R., "Constructing Object Models from Multiple Images", Intl. J. of Computer Vision, 9, p185-212, 1992.
No context found.
J.R. Stenstrom and C.I. Connolly. Constructing object models from multiple images, International Journal of Computer Vision, 9(3):185--212, 1992.
No context found.
J.R.Steenstrom and C.I.Connolly, `Constructing object models from multiple images', Int. J. of Computer Vision, 9 (1992), 185-212.
No context found.
J. Stenstrom and C. I. Connolly. Constructing object models from multiple images. Int. Journal of Computer Vision, 9(3):185--212, 1992.
No context found.
Stenstrom, J.R. and Connolly, C.I.: Constructing Object Models from Multiple Images, International Journal of Computer Vision, Vol. 9, No. 3, pp. 185212, 1992
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC