Philip H. S. Torr and Andrew Zisserman. Feature based methods for structure and motion estimation. In Workshop on Vision Algorithms, 1999.  Home/Search   Document Details and Download   Summary   Related Articles   Check

.... mappings such as translational, rotational, affine or homographic mappings [2 4] They have also been more complex and involved optical flow calculations [5] and the modeling of lens distortion [6] These domain transformations are typically determined using either direct [7] or feature based [8] methods. The registration of images tonally has seen work that models a camera s response function so that multiple images of a static scene, captured at different exposure settings, can been used to create high dynamic range maps whose resolution is greater than the typical 8 bits pixel channel ....

P. Torr and A. Zisserman, "Feature Based Methods for Structure and Motion Estimation," International Workshop on Vision Algorithms, pp. 278 -- 295, 1999.

....results are provided on large image sequences. In particular, we consider sport videos where the moving and deforming athlet is visible in every frame of the sequence, thereby making tricky the alignment task. 1. Introduction Panoramic photography has received a growing interest since a decade [2, 3, 7, 8, 11, 13, 15]. It consists in stitching images to form wide angle mosaics. Among their numerous applications, such techniques may be used to efficiently represent video sequences, in terms of compression, enhancement, vizualization etc, see [8] A number of papers, see e.g. 11] concentrate on the static ....

....fine when the dynamic areas are small compared to the scene size. The most important features of the techniques that may be used to construct such mosaics are the ability to handle large dynamic layers, and the accuracy in frame alignment. Two classes of techniques can be used: feature based [13] and direct [7] methods. Each of them has specific advantages and drawbacks. In particular, we observed that featurebased methods are more robust, in terms of outling features (those who lie on e.g. the dynamic layer and therefore, who do not fulfill the motion model) while direct methods are more ....

[Article contains additional citation context not shown here]

P. H. S. Torr and A. Zisserman. Feature based methods for structure and motion estimation. In Vision Algorithms: Theory and Practice, July 1999.

....image is compared to the background panorama in order to extract image regions corresponding to motion the dynamic layer. Finally, the static and dynamic layers are combined together to form a motion panorama (bottom) Panoramic photography has received growing interest since a decade [2, 3, 11, 12, 20, 22, 25, 15, 16] resulting in a number of commercial products such as [18] The idea behind these methods is that there exist a simple invertible transformation between images gathered with a camera rotating around its center of projection [9] A vast majority of papers (see [20] for a review) concentrates on ....

....images. 3 background. The most crucial characteristics of methods associated with motion panorama construction are (i) the ability to deal with large dynamic image regions and (ii) the accuracy in frame alignment. Generally speaking, two categories of methods are available: Featurebased methods [22] and direct methods [11] The former consists in extracting image features such as points of interest, matching such features over several images, and estimating the mapping between images based on feature to feature correspondences. The latter consists of finding the image to image mapping which ....

P. H. S. Torr and A. Zisserman. Feature based methods for structure and motion estimation. In W. Triggs, A. Zisserman, and R. Szeliski, editors, Vision Algorithms: Theory and Practice, number 1883 in LNCS, pages 278--295, Corfu, Greece, July 1999. Springer-Verlag.

....reality [1] mobile robot navigation [2] face tracking [3] or the generation of super resolution images [4] Traditional approaches to tracking are based on nding correspondences in successive images. This can be achieved by computing optical ow [5] or by matching a sparse collection of features [6]. In ow based methods, a velocity vector is computed for each pixel, while in feature based methods, image features, such as points and lines are matched across all frames in the sequence. Feature based methods minimise an error measure based on geometrical constraints between a few corresponding ....

P. H. S. Torr and A. Zisserman, "Feature based methods for structure and motion estimation," in Vision Algorithms: Theory and practice, W. Triggs, A. Zisserman, and R. Szeliski, Eds. Springer-Verlag, 1999, pp. 278-- 295.

....[5] mobile robot navigation [7] face tracking [1] or the generation of super resolution images [2] Traditional approaches to tracking are based on finding correspondences in successive images. This can be achieved by computing optical flow [4] or by matching a sparse collection of features [6]. In flow based methods, a velocity vector is computed for each pixel, while in feature based methods, image features, such as points and lines are matched across all frames in the sequence. Feature based methods minimise an error measure based on geometrical constraints between a few ....

P. H. S. Torr and A. Zisserman. Feature based methods for structure and motion estimation. In W. Triggs, A. Zisserman, and R. Szeliski, editors, Vision Algorithms: Theory and practice, pages 278--295. Springer-Verlag, 1999.

No context found.

P. H. S. Torr and A. Zisserman. Feature based methods for structure and motion estimation. In W. Triggs, A. Zisserman, and R. Szeliski, editors, Vision Algorithms: Theory and Practice, volume 1883 of LNCS, pages 278--294. Springer, 2000. 10

....of the distance is essentially that of estimating the affine transformation which best aligns the two images, and reporting the squared error between the aligned and original images. This is a problem with a long history, and two basic approaches may be discerned: direct [13] and feature based [29]. The distance metric in this paper is a purely direct method without any feature correspondences. 3.1 AIDM: Affine invariant distance measure In general, the transformation will not be linear in the parameters a, so we shall linearize about the origin. Expanding in terms of derivatives of I ....

P. H. S. Torr and A. Zisserman. Feature based methods for structure and motion estimation. In W. Triggs, A. Zisserman, and R. Szeliski, editors, Vision Algorithms: Theory and Practice, volume 1883 of LNCS, pages 278--294. Springer, 2000.

No context found.

Philip H. S. Torr and Andrew Zisserman. Feature based methods for structure and motion estimation. In Workshop on Vision Algorithms, 1999.

No context found.

Philip H. S. Torr and Andrew Zisserman. Feature based methods for structure and motion estimation. In Workshop on Vision Algorithms, pages 278--294, 1999.

No context found.

P. H. S. Torr and A. Zisserman. Feature based methods for structure and motion estimation. In W. Triggs, A. Zisserman, and R. Szeliski, editors, Vision Algorithms: Theory and practice, pages 278--295. Springer-Verlag, 1999.

No context found.

P. H. S. Torr and A. Zisserman, "Feature based methods for structure and motion estimation," in Vision Algorithms: Theory and practice, W. Triggs, A. Zisserman, and R. Szeliski, Eds. Springer-Verlag, 1999, pp. 278--295.

No context found.

Philip H. S. Torr and Andrew Zisserman. Feature based methods for structure and motion estimation. In Workshop on Vision Algorithms, pages 278--294, 1999.

No context found.

Philip H. S. Torr and Andrew Zisserman. Feature based methods for structure and motion estimation. In Workshop on Vision Algorithms, pages 278--294, 1999.

No context found.

P. Torr and A. Zisserman, "Feature based methods for structure and motion estimation", in Int. Work. on Vis. Alg., p. 278-295, 1999.

No context found.

P. Torr and A. Zisserman. Feature Based Methods for Structure and Motion Estimation. In B. Triggs, A. Zisserman, and R. Szeliski, editors, Vision Algorithms: Theory and Practice, number 1883 in LNCS, pages 278--294. Springer-Verlag, Corfu, Greece, September 1999.

No context found.

P. H. S. Torr and A. Zisserman. Feature based methods for structure and motion estimation. In W. Triggs, A. Zisserman, and R. Szeliski, editors, International Workshop on Vision Algorithms: Theory and Practice, pages 278--295. Springer Verlag, 1999.

No context found.

P. Torr and A. Zisserman. Feature Based Methods for Structure and Motion Estimation, Vision Algorithms: Theory and Practice, pp. 278-294, Corfu, Greece, 1999.

No context found.

P. Torr and A. Zisserman. Feature Based Methods for Structure and Motion Estimation. In B. Triggs, A. Zisserman, and R. Szeliski, editors, Vision Algorithms: Theory and Practice, number 1883 in LNCS, pages 278--294. Springer-Verlag, Corfu, Greece, September 1999.

No context found.

P. H. S. Torr and A Zisserman. Feature based methods for structure and motion estimation. InW. Triggs, A. Zisserman, and R. Szeliski, editors, International Workshop on Vision Algorithms, pages 278--295, 1999.

No context found.

P. H. S. Torr and A. Zisserman. Feature based methods for structure and motion estimation. In B. Triggs, A. Zisserman, and R. Szelisky, (eds.), Vision Algorithms: Theory and Practice, pages 278--294. Springer, 2000.

No context found.

P. Torr and A. Zisserman, "Feature based methods for structure and motion estimation," in Proc. Int. Workshop on Vision Algorithms, 1999, pp. 278--295.

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