S. Laveau and O. Faugeras, "3-D Scene Representation as a Collection of Images and Fundamental Matrices," Technical Report RR-2205, INRIA, 1994.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....issues in novel view generation. We present results on rendering of realistic walk through sequences for both indoor and outdoor scenes which demonstrate the applicability of our method. 1. Introduction Recently, the problem of image based rendering has attracted considerable attention [2, 8, 12, 14, 4, 1], wherein the environmental map for rendering of novel views are maintained in terms of a set of images instead of explicit geometric and photometric models of a scene. In this paper we address the problem of generation of a sequence of novel views of a scene from a set of reference views. We ....

....point in a target image, and holes, where a region occluded in a reference image becomes visible in a target image. Our method is based on segmentation of the scene in terms of homographies between regions in the reference images through the planar patches. 1.1. Previous work Laveau and Faugeras [8] present a method of representing a 3D scene as a collection of reference images and their pairwise epipolar geometries. They propose a 2D ray tracing algorithm which start from each point in the target image and locate the corresponding points in the reference images, from which the intensity ....

[Article contains additional citation context not shown here]

S. Laveau and O. D. Faugeras, "3-D Scene Representation as a Collection of Images and Fundamental Matrices," INRIA, Technical Report No. 2205, 1994.

....Unlike traditional polygon systems in which rendering time is proportional to scene complexity, in image based systems rendering time is proportional to the number of pixels in the source images or the output image. The techniques of view interpolation [Chen93, Chen95] epipolar geometry [Laveau94], plenoptic modeling [McMillan95] light fields [Levoy96] lumigraphs [Gortler96] view based rendering [Pulli97] nailboards [Schaufler97] and post rendering 3D warping [Mark97] have all demonstrated the speed and viability of image based graphics. Despite the success of these approaches, they ....

....continuous motion between viewpoints by morphing from one to the next. However, the virtual camera was constrained to lie on the lines joining adjacent viewpoints, and the linear interpolation of the optical flow vectors which was used for morphing provided approximate view reconstruction only. In [Laveau94] and [McMillan95] the optical flow information was combined with knowledge of the relative positions of the cameras used to acquire the images. This allowed the researchers to provide exact view reconstruction from arbitrary viewpoints. The idea of using images to represent single objects (as ....

[Article contains additional citation context not shown here]

S. Laveau, O.D. Faugeras, "3-D Scene Representation as a Collection of Images and Fundamental Matrices", INRIA Technical Report No. 2205, February 1994

....viewpoints and use these as an environment map [6] to be applied as a texture on some imaging surface. In this paper we are interested in actually generating new views from a small set of existing ones. This problem, called view synthesis, has received considerable recent interest [10] 12] 5] [9], 1] 4] 3] 7] 14] because, although in its simplest form it is better replaced by a short video sequence (a spatially dense version of [6] its potential applications are much more far reaching. For instance we have experimented with applying the method to image sequences in order to obtain a ....

....alternative views. Chen and Williams [5] have also studied the interpolation of intermediate views in the context of 3D graphics using the British Machine Vision Conference 771 perfect dense correspondences implicit in the synthetic data. In the second category is the work of Laveau and Faugeras. [9] and more recently Avidan and Shashua [1] which use projective transfer to predict, from dense correspondences, where pixels end up in virtual projectively distorted images 1 . Besides requiring hard to obtain dense correspondences, approaches in either of the first two categories suffer from ....

[Article contains additional citation context not shown here]

Laveau, S & Faugeras, O., "3D Scene Representation as a Collection of Images and Fundamental Matrices", INRIA Tech Report 2205, February 1994.

....environment map from different vantage points to warp between multiple images. The key challenge in this warping approach is to fill in the gaps when previously occluded areas become visible. Another approach to interpolating between acquired images is to find corresponding points in the two [Laveau94, McMillan95b, Seitz95]. If the positions of the cameras are known, this is equivalent to finding the depth values of the corresponding points. Automatically finding correspondences between pairs of images is the classic problem of stereo vision, and unfortunately although many algorithms exist, these algorithms are ....

Laveau, S., Faugeras, O.D., "3-D Scene Representation as a Collection of Images and Fundamental Matrices," INRIA Technical Report No. 2205, 1994.

....also the key to establishing further correspondences in image matching. Luong and Faugeras (1996) discuss many other applications, including the construction of stereo projective invariants (Gros and Quan 1992, Hartley 1993a, Hartley 1993b, Shashua 1993) the synthesis of new images from old ones (Laveau and Faugeras 1994, Seitz and Dyer 1995b) convex hull computation (Robert and Faugeras 1993) image rectification (Hartley and Gupta 1993, Seitz and Dyer 1995a) motion segmentation (Nishimura et al. 1993, Torr and Murray 1993) and self calibration (Faugeras et al. 1992, Hartley 1994) As mentioned by Faugeras ....

Laveau, S. and Faugeras, O.: 1994, 3D scene representation as a collection of images and fundamental matrices, Technical Report 2205, INRIA Sophia-Antipolis.

.... spherical frames can be determined without an explicit representation of the range information, or , as shown below: Unlike range information, the generalized disparity term, can be determined directly from observable image features, such as optical flow [Prazdny83] or point correspondences [Laveau94]. These two dimensional mapping functions, may result in several points from the domain mapping to a single point in the range. Such range points will be referred to as topological multiplicities of the mapping. The specific claim of this report is that an occlusion compatible mapping can also ....

Laveau, S. and O. Faugeras, "3-D Scene Representation as a Collection of Images and Fundamental Matrices," INRIA, Technical Report No. 2205, February, 1994.

....PYI NSF grant MIP 9057466, ONR Young Investigator Award N00014 92 J 1732, and Sun Microsystems. Note that this manuscript differs slightly from the published version. Please contact IEEE directly for a copy of the actual paper Publisher Item Identifier S 1057 7149(97)02463 9. problem [38] 33] [25]. This approach however correctly reconstructs only those points that lie in the intersection of the given views and not points that become uncovered. In the case of [25] despite being able to generate novel views from a set of reference views, one requires dense correspondences and the epipolar ....

....contact IEEE directly for a copy of the actual paper Publisher Item Identifier S 1057 7149(97)02463 9. problem [38] 33] 25] This approach however correctly reconstructs only those points that lie in the intersection of the given views and not points that become uncovered. In the case of [25], despite being able to generate novel views from a set of reference views, one requires dense correspondences and the epipolar geometry to be known a priori . In addition, these kinds of approach would require a considerable amount of computation to continuously update a user defined viewpoint ....

[Article contains additional citation context not shown here]

S. Laveau and O. Faugeras, "3-D scene representation as a collection of images and fundamental matrices," Tech. Rep. 2205, INRIA, Feb. 1994.

....methods. The limitation of texture maps is that the image elements are constrained to follow the underlying geometry of the primitive upon which the texture is mapped. The recently introduced methods of view interpolation [Chen93] Quicktime VR [Chen95] fundamental matrixbased re projection [Laveau94], and plenoptic modeling [McMillan95b]demonstrate the potential of the image based rendering approach. All of these methods take advantage of the incremental evaluation in their image warps and the amount of computation required is proportional to either the number of pixels in the reference ....

S. Laveau and O. Faugeras, "3-D Scene Representation as a Collection of Images and Fundamental Matrices," INRIA, Technical Report No. 2205, February 1994.

....interpolation of corresponding image points using range data to obtain correspondences. They investigated special situations in which interpolation produces valid perspective views, but concluded that the interpolated images do not in general correspond to exact perspective views. Two groups [2, 6] have recently developed image warping techniques for perspective correct view synthesis. Under the assumption that a complete pixelwise correspondence is available, it is possible to predict a broad range of views. Several researchers [1, 7, 8, 9, 10] have used interpolation to produce new images ....

S. Laveau and O. Faugeras, "3-D scene representation as a collection of images and fundamental matrices," Tech. Rep. 2205, INRIA, SophiaAntipolis, France, February 1994.

....motion. We propose to represent the scene by dense depth maps and corresponding intensity information at selected viewpoints in the scene called reference frames. Previous approaches attempt to construct a full 3 D model (e.g. 1] or reconstruct views directly without first estimating depth (e.g. [2]) More details are given in [3] Our approach with dense depth maps is not as complex as 3 D modeling and is capable of handling occlusions better than the direct methods. 2. COMPACT REPRESENTATION Our overall approach in deriving depth information at reference locations is to establish ....

S. Laveau and O. Faugeras, "3-D scene representation as a collection of images and fundamental matrices," Tech. Rep. 2205, INRIA, Feb. 1994.

....Unlike traditional polygon systems in which rendering time is proportional to scene complexity, in image based systems rendering time is proportional to the number of pixels in the source images or the output image. The techniques of view interpolation [Chen93, Chen95] epipolar geometry [Laveau94], plenoptic modeling [McMillan95] light fields [Levoy96] lumigraphs [Gortler96] view based rendering [Pulli97] nailboards [Schaufler97] and post rendering 3D warping [Mark97] have all demonstrated the speed and viability of image based graphics. Despite the success of these approaches, ....

S. Laveau, O.D. Faugeras, "3-D Scene Representation as a Collection of Images and Fundamental Matrices", INRIA Technical Report No. 2205, February 1994

No context found.

S. Laveau and O. Faugeras, "3-D Scene Representation as a Collection of Images and Fundamental Matrices," Technical Report RR-2205, INRIA, 1994.

No context found.

Stpphane Laveau, Olivier Faugeras, "3-D Scene Representation as a Collection of Images and Fundamental Matrices", Technical Report 2205, INRIA, 1994.

No context found.

S. Laveau, O.D. Faugeras, "3-D Scene Representation as a Collection of Images and Fundamental Matrices", INRIA Technical Report No. 2205, February 1994

No context found.

S. Laveau and O. Faugeras, "3-D Scene Representation as a Collection of Images and Fundamental Matrices," INRIA, Technical Report No. 2205, February, 1994.

No context found.

Laveau, S. and O. Faugeras, "3-D Scene Representation as a Collection of Images and Fundamental Matrices," INRIA, Technical Report No. 2205, February, 1994.

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