M. Okutomi and T. Kanade. A multiple-baseline stereo. IEEE Transactions on Pattern Analysis and Machine Intelligence, 15(4):353--63, 1993.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....6 The two previous mentioned approaches to surface reconstruction illustrate a clear and well known dichotomy in baseline stereo: smaller motions aid with the correspondence problem, while greater baseline motion provide more stable depth estimates but sparser data. Multi baseline stereo [2, 37, 44, 53, 69] has sought to merge these two problems by tracking many points over many small motions. Once a sufficiently large baseline has been achieved, depth values are computed with less ambiguity. As mentioned earlier, this approach is not practical for an autonomous explorer that does not necessarily ....

Okutomi Kanade, A Multiple-Baseline Stereo, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 15, No. 4, pp. 355-363, April 1993.

....to this problem have been proposed in the literature, ranging from a hierarchical smoothing or coarse to fine strategy, to a global optimization technique based on surface coherence assumptions. These techniques, however, tend to be heuristic or result in computationally expensive algorithms. In [57], Okutomi and Kanade propose an efficient multi baseline stereo vision method, which is more accurate for depth estimation than more conventional approaches. Whereas, in ordinary stereo, depth 1 input image 5 input images 1 template 10 templates 1 template 10 templates # CPUs (s) s) s) s) 1 ....

Okutomi M, Kanade T. A multiple-baseline stereo. IEEE Transactions on Pattern Analysis and Machine Intelligence 1993; 15(4):353--363.

....function in the other image. 1.1.2 Aggregation method The aggregation method refers to the manner in which the error function over the search space is computed or accumulated. The most direct way is to apply search windows of a fixed size over a prescribed disparity space for multiple cameras [OK93] or for verged camera configuration [KWZK95] Others use adaptive windows [OK92] shiftable windows [Arn83, BI99, TSK01] or multiple masks [NMSO96] Another set of methods accumulates votes in 3D space, e.g. the space sweep approach [Col96] and voxel coloring and its variants [SD97, SG99, KS99] ....

....k ,k=0. K . I 0 (x, y) is the reference image for which we wish to compute a disparity map d(x, y) such that pixels in I 0 (x, y) project to their corresponding locations in the other images when the correct disparity is selected. In the classic forward facing multi baseline stereo configuration [OK93], the camera matrices are such that disparity (inverse depth) varies linearly with horizontal pixel motion, I k (x, y, d) I k (x b k d(x, y) y) 1) I k (x, y, d) is image I k warped by the disparity map d(x, y) In a more general (plane sweep) multi view setting [Col96, SG99] each ....

M. Okutomi and T. Kanade. A multiple baseline stereo. IEEE Transactions on Pattern Analysis and Machine Intelligence, 15(4):353--363, April 1993.

....partly reduced by using better matching cost measures, which are more discriminable. On the other hand, the structural ambiguity can be significantly reduced by using more cameras observing the scene from various points of view, because the polynocular case de ambiguate the correspondence problem [48, 13, 52, 35]. In our research (and consequently in this survey as well) we focus on binocular stereo matching, i.e. the matching problem is solved using only two input images. To be able to compute correspondences using only two images correctly is highly thorny challenge. Reduction of the number of cameras ....

....because it straightforwardly determines the quality of results. Consequently, it is very important to design a method based on a good logical basis, not just on some heuristics which works well . There exist di#erent principles for selecting the matches, such as winner takes all (WTA) [48, 65], or Stable matching [54, 55] In the WTA approach, for each point in a reference image is (somehow) selected the best tentative pair (the winner) However, in Stable matching, the symmetric best unique solution is identified. In order to produce accurate results, selection methods require to ....

[Article contains additional citation context not shown here]

Masatoshi Okutomi and Takeo Kanade. A multiple-baseline stereo. IEEE Transactions on Pattern Analysis and Machine Intelligence, 15(4):353--363, 1993.

....(SSD) over a 1 1 window in the second camera. Typically, the SSD score over a single pixel support window does not provide sufficient disambiguating power, so it is essential for our method to use several cameras. As such our method can be considered in effect a multi baseline stereo algorithm [25] operating with a 1 1 support window, with the goal being the best color estimate rather than depth. Conceptually, our method distributes the correlation support among input cameras. In binocular stereo, the support is distributed around the neighborhood of the pixel of interest within a ....

M. Okutomi and T. Kanade. A Multi-baseline Stereo. IEEE Transactions on Pattern Analysis and Machine Intelligence, 15(4):353--363, April 1993.

....the classical technique of dynamic programming [2] Similar approaches have been often used in computer vision. Our contribution is to work directly in the 3D space and not in the disparity space as [11, 12, 16] so we have a more robust localization against ambiguity as shown by Okutomi and Kanade [17]. We also use dynamic programming only in the regions known to represent a smooth surface, unlike other techniques [6, 7, 11, 12, 16, 17] that work with the whole scan line and then need some special treatments to cope with occlusions or discontinuities. In addition we introduce the vertex ....

.... is to work directly in the 3D space and not in the disparity space as [11, 12, 16] so we have a more robust localization against ambiguity as shown by Okutomi and Kanade [17] We also use dynamic programming only in the regions known to represent a smooth surface, unlike other techniques [6, 7, 11, 12, 16, 17] that work with the whole scan line and then need some special treatments to cope with occlusions or discontinuities. In addition we introduce the vertex coordinate property to characterize rigorously the search area. Figure 7 shows an epipolar plane : on the left, the standard deviation is ....

M. Okutomi and T. Kanade. A multiple-baseline stereo. IEEE Transactions on Pattern Analysis and Machine Intelligence, 15(4):353--63, 1993.

....of the task parallel framework based on the IATG we compare the speed ups obtained by applying only data parallelism to the application, with the speed ups obtained with both data and task parallelism. The multi baseline stereo vision application uses an algorithm developed by Okutomi and Kanade [9] and described by Webb and al. 18, 17] that gives greater accuracy in depth through the use of more than two cameras. Input consists of three TyxzT images acquired from three horizontally aligned, equally spaced cameras. One image is the reference image, the other two are named match images. ....

M. Okutomi and T. Kanade. A mutiple-baseline stereo. IEEE Transactions on Pattern Analysis and Machine Intelligence, 15(4):353--363, 1993.

....don t always converge and don t always recover the correct correspondences. Another approach is to use multiple images. Several researchers, such as Yachida [11] have proposed trinocular stereo algorithms. Others have also used special camera configurations to aid in the correspondence problem, [10, 1, 8]. Bolles, Baker and Marimont [1] proposed constructing an epipolarplane image from a large number of images. In some cases, analyzing the epipolar plane image is much simpler than analyzing the original set of images. The epipolar plane image, however, is only defined for a limited set of camera ....

....constructing an epipolarplane image from a large number of images. In some cases, analyzing the epipolar plane image is much simpler than analyzing the original set of images. The epipolar plane image, however, is only defined for a limited set of camera positions. Tsai [10] and Okutomi and Kanade [8] defined a cost function which was applied directly to a set of images. The extremum of this cost function was then taken as the correct correspondence. Occlusion is assumed to be negligible. In fact, Okutomi and Kanade state that they invariably obtained better results by using relatively short ....

[Article contains additional citation context not shown here]

Masatoshi Okutomi and Takeo Kanade. A muliple-baseline stereo. IEEE Transactions on Pattern Analysis and Machine Intelligence, 15(4):353--363, April 1993.

....recover a full depth map, i.e. to infer the missing component associated with each pixel in a given image sequence. When the camera motion is known, the problem of depth map recovery is called stereo reconstruction (or multi frame stereo if more than two views are used [Matthies et al. 1989; Okutomi and Kanade, 1993]. This problem has been extensively studied in photogrammetry and computer vision [Moffitt and Mikhail, 1980; Barnard and Fischler, 1982; Horn, 1986; Dhond and Aggarwal, 1989] When the camera motion is unknown, we have the more difficult structure from motion problem [Horn, 1986; Weng et al. ....

M. Okutomi and T. Kanade. A multiple baseline stereo. IEEE Transactions on Pattern Analysis and Machine Intelligence, 15(4):353--363, April 1993.

....of the task parallel framework based on the IATG we compare the speed ups obtained by applying only data parallelism to the application, with the speed ups obtained with both data and task parallelism. The multi baseline stereo vision application uses an algorithm developed by Okutomi and Kanade [9] and described by Webb and al. 18, 17] that gives greater accuracy in depth through the use of more than two cameras. Input consists of three n n images acquired from three horizontally aligned, equally spaced cameras. One image is the reference image, the other two are named match images. For ....

M. Okutomi and T. Kanade. A mutiple-baseline stereo. IEEE Transactions on Pattern Analysis and Machine Intelligence, 15(4):353--363, 1993.

No context found.

M. Okutomi and T. Kanade. A multiple-baseline stereo, IEEE Transactions on Pattern Analysis and Machine Intelligence. 15(4):353-363, 1993.

No context found.

M. Okutomi and T. Kanade. A Multiple-Baseline Stereo. IEEE Transactions on Pattern Analysis and Machine Intelligence, 15(4):353--363, April 1993.

No context found.

M. Okutomi and T. Kanade. A multiple-baseline stereo. IEEE Transactions on Pattern Analysis and Machine Intelligence, 15(4):353--363, 1993.

....pixels. The camera coordinate frames are co planar and aligned so that the the epipolar lines of the camera pairs lie along the rows and columns of the images. 2. 2 Trinocular Stereo Algorithm The trinocular stereo approach is based on the multibaseline stereo developed by Okutomi and Kanade [14]. Each pixel in the reference image is compared with pixels along the epipolar lines in the top and left images. The comparison measure used is sum of absolute differences (SAD) The results of the two image pairs (left right, top bottom) are summed to yield a combined score. Multibaseline stereo ....

M. Okutomi and T. Kanade. A multiple-baseline stereo. IEEE Transactions on Pattern Analysis and Machine Intelligence, 15(4):353--363, 1993.

....points or patches is critical, but often difficult and expensive. One well known constraint for reducing these ambiguities is to add a third camera to verify hypothesized matches. However, most trinocular systems proposed in the literature exploit a right triangular [5, 1, 3, 4] and or parallel [6] camera triple configuration. The close range scanning tasks we are interested in are better served by a surround configuration of cameras, which disallows the triple rectification methods which simplify these parallel camera approaches. The trinocular epipolar constraint in stereo vision is ....

.... 1, 3] This correspondence and verification is simplified if the cameras are aligned or rectified in an up down left right parallel right triangle configuration, where epipolar lines are made parallel to the horizontal and vertical scanlines [1] For dense disparity calculation, Okutomi and Kanade [6] used a linear parallel configuration of cameras in their multibaseline system. They exploited a third view by summing pairwise SSD values referred to common 1 Z values. R dr C L b b C C R L L R Figure 1. Trinocular camera triple. The configurations we are interested in are similar ....

[Article contains additional citation context not shown here]

M. Okutomi and T. Kanade. A multiple-baseline stereo. IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), 15(4):353--363, April 1993.

....This high performance combined with wide angle lenses allows the robot to see a large area quickly so that it can map without pausing and move at an acceptable velocity. The stereo system computes depth maps using three calibrated cameras with an algorithm similar to the multi baseline stereo [10]. On Spinoza the first pair of cameras are in a horizontal plane while the second are in the vertical plane. The trinocular cameras can achieve better results than a typical two camera stereo system because the second pair of cameras can resolve situations that are ambiguous to the first pair. For ....

M. Okutomi and T. Kanade. A multiple--baseline stereo. IEEE Transactions on Pattern Analysis and Machine Intelligence, PAMI--15(4):353--363, April 1993.

....2, placed cameras on an arc surrounding the user at the same level. This does not allow us to arrange or rectify triples of camera image planes such that they are coplanar, and therefore it is more expensive for us to exploit the trinocular constraint. Following Okutomi and Kanade s observation [21], we optimize over the sum of correlation values with respect to the true depth value rather than disparity. Essentially we treat the camera triple #L, C, R# as two independent stereo pairs #L, CL # and #CR , R#. When revising our system design to parallelize and improve its speed, we ....

Masatoshi Okutomi and Takeo Kanade. A multiplebaseline stereo. IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), 15(4):353--363, April 1993.

No context found.

M. Okutomi and T. Kanade. A multiple-baseline stereo. IEEE Transactions on Pattern Analysis and Machine Intelligence, 15(4):353--63, 1993.

No context found.

M. Okutomi and T. Kanade. A Multi-baseline Stereo. IEEE Transactions on Pattern Analysis and Machine Intelligence, 15(4):353--363, April 1993.

No context found.

M. Okutomi and T. Kanade. A multiple-baseline stereo. IEEE Transactions on Pattern Analysis and Machine Intelligence, 15:353--363, 1993.

No context found.

M. Okutomi and T. Kanade. A Multiple-Baseline Stereo. IEEE Transactions on Pattern Analysis and Machine Intelligence, 15(4):353--363, April 1993.

No context found.

M. Okutomi and T. Kanade. A multiple-baseline stereo. IEEE Transactions on Pattern Analysis and Machine Intelligence, 15(4):353--363, 1993. 1

No context found.

M. Okutomi and T. Kanade. A multiple-baseline stereo. ieee Transactions on Pattern Analysis and Machine Intelligence, 15(4):353-363, April 1993.

No context found.

M. Okutomi and T. Kanade. A multiple baseline stereo. IEEE Transactions on Pattern Analysis and Machine Intelligence, 15(4):353--363, April 1993.

No context found.

Okutomi M. and Kanade, T.: A Multiple-Baseline Stereo. IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 15, No. 4, pp. 355-363, April 1993.

First 50 documents  Next 50

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