Perception and Action using Multilinear Forms

Size: px
Start display at page:

Download "Perception and Action using Multilinear Forms"

Transcription

1 Perception and Action using Multilinear Forms Anders Heyden, Gunnar Sparr, Kalle Åström Dept of Mathematics, Lund University Box 118, S Lund, Sweden Abstract In this paper it is shown how multilinear forms can be used in the perception-action cycle. Firstly, these forms can be used to reconstruct an unknown (or partially known) scene from image sequences only. Secondly, from this reconstruction the movement of the camera can be calculated with respect to the scene, which solves the so called hand-eye calibration problem. Then action can be carried out when this relative orientation is known. The results are that it is sufficient to either use bilinear forms between every successive pair of images plus bilinear forms between every second image or trilinear forms between successive triplets of images. We also present a robust and accurate method to obtain reconstruction and hand-eye calibration from a sequence of images taken by uncalibrated cameras, based on multilinear forms. This algorithm requires no initialisation and gives a generic solution in a sense that is clearly specified. Finally, the algorithms are illustrated using real image sequences, showing that reconstruction can be made and that the hand-eye calibration is obtained. 1 Introduction The problem of reconstruction of an unknown scene from a number of its projective images has been studied by many researchers. Firstly, calibrated cameras where considered, see [13], making it possible to reconstruct the scene up to an unknown similarity transformation (Euclidean plus scale). The drawback of this approach is that the camera need to be calibrated before the image sequence is captured and when this have been done it is neither possible to change focus nor to zoom. Another drawback is that the algebra gets complicated because orthogonal matrices have to be used, giving algebraic constraints on each such matrix. Another approach is to model the projective transformation as an affine transformation, giving an affine reconstruction of the scene, see [21]. This approach gives a very simple algebraic formulation and the reconstruction and hand-eye calibration can be obtained using linear methods. The drawback is that the approximation, giving an affine camera, is often not accurate, e.g. when there are perspective effects present. When some action has to be made, the camera often has to move closer to the scene, giving large perspective effects, which makes this approximation very bad in perception-action tasks. Other similar attempts are orthographic projections, see [21] and paraperspective, see [15]. This work has been supported by the Swedish Research Council for Engineering Sciences (TFR), project

2 Recently, methods requiring no camera calibration have become popular, [2,14,19]. These methods are based on projective geometry and give a reasonable algebra. The first result obtained is that it is only possible to reconstruct the object up to an unknown projective transformation, see [18,2,14,5]. This drawback is not always important, especially in perception-action tasks, because this kind of reconstruction, together with the head-eye calibration is often sufficient in order to take some action, e.g. obstacle avoidance, positioning, picking up objects, navigation, etc. The first algorithms used the so called fundamental matrix, see [2], which is obtained from the bilinear form. Later on, higher order multilinear forms where introduced; trilinearities between triplets of views, see [16,6,3], and quadrilinearities between quadruples of views, see [22,8,3]. It soon became apparent that there was no need to go beyond quadrilinearities and after a while it was discovered that the trilinearities was sufficient to use. Recently, the algebraic and geometric properties of the bilinearities and the trilinearities have been exploited, see [10], and it has been shown that both geometrically and algebraically, the bilinearities are sufficient to use. These multilinear forms can be used both to reconstruct the scene and to obtain hand-eye calibration. There has also been intensive research on more robust methods, based on the multilinear forms, see [23,20,11]. These iterative methods requires no initialisation step and gives optimal reconstruction in a sense that will be specified later. The advantage of these methods is that they use all images and all available point matches with equal priority, i.e. no image are selected as a reference image. They can easily be extended to recursive algorithms, taking more and more images into account as they become available, see [1]. In this paper, we will present reconstruction methods based on multilinear forms and iterative robust methods. It will be shown, both theoretically and in experiments that an accurate reconstruction and hand-eye calibration are obtained. In this way the perception-action problem is treated. 2 Camera Model The image formation system (the camera) is modeled by the equation (1) Here denotes object coordinates in extended form and denotes extended image coordinates. The scale factor, called the depth, accounts for perspective effects and represents a rigid transformation of the object, i.e. denotes a! rotation matrix and a! translation vector. Finally, the parameters in represent intrinsic properties of the image formation system: and represent magnifications in the - and -directions in the light sensitive area, represents the skew, i.e. nonrectangular arrays can be modelled, and " # $ is called the principal point and is interpreted as the orthogonal projection of the focal point onto the image plane. The parameters in and are called extrinsic parameters and the parameters in

3 are called the intrinsic parameters. Observe that there are extrinsic and intrinsic parameters, totally, the same number as in an arbitrary! matrix defined up to a scale factor. If the extrinsic as well as the intrinsic parameters are unknown (1) can compactly be written (2) Since there is a freedom in the choice of coordinate system in the object, it is a general practice to chose, which will be done in the sequel. However, three degrees of freedom remains, since a projective change of coordinates by the matrix does not change. This ambiguity in the determination of the camera matrices corresponds to the fact that it is only possible to reconstruct the scene up to an unknown projective transformation. This ambiguity in representation will be called the projective ambiguity. It is furthermore convenient to use the notation # # (3) for the first three camera matrices. In the sequel we will assume that we have points (with known correspondences) in different images and that the intrinsic parameters are allowed to vary between the different imaging instants. Image coordinates in image are denoted by and the camera matrix for image number will be denoted by, i.e. # # # (4) 3 Multilinear Forms Consider the equations in (4). These equations can be written # (5) with #. (6) Since has a nontrivial nullspace, it follows that " $ (7)

4 The matrix in (6) contains one block with three rows for each image. Observe that all determinants of " $! " $ submatrices of in (6) are multihomogeneous of degree " # # # # $, that is of the same degree in every triplet of image coordinates. Definition 1. The subdeterminants of size " $! " $ from in (6) are called the multilinear constraints. The multilinear constraints obtained from submatrices containing all rows corresponding to two images and one row from each of the other images are called the bilinear constraints. The bilinear constraints between image and image can be written as a product of -, - and-coordinates in the other images and " # # # # #$ (8) Since the first factors consists of projective coordinates, some combination of these projective coordinates has a nonvanishing product and the bilinear constraints are equivalent to the constraint in (8), which is sometimes called the epipolar constraint. These constraints can be written # (9) where is called the fundamental matrix between images and. Given at least corresponding points in images, it is possible to solve linearly for the entries in by using 9. Then it is possible to calculate the camera matrices and up to the projective ambiguity. The multilinear constraints obtained from submatrices containing all rows corresponding to one image, two rows each from two other images and one row from each of the other images are called the trilinear constraints. The trilinear constraints between image, and can be written as a product of -, - and -coordinates in the other images and! subdeterminants of (10) Again the first factors consists of projective coordinates, and some combination of these projective coordinates has a nonvanishing product, thus the trilinear constraints are equivalent to the constraints expressed by subdeterminants from (10). The trilinear constraints are often expressed using the so called trilinear tensor,, with components defined by " $ " $ # (11) using the notations in (3), see [17,6,8]. There turns out to be four linearly independent (in ) trilinear constraints, obtained from subdeterminants of (10). This makes it possible to obtain four linear constraints on the coefficients of the trilinear tensor for

5 each corresponding point in images. Thus having corresponding points in images makes it possible to linearly recover the components of the trilinear tensor and from them the camera matrices (up to the previous mentioned ambiguity). The multilinear constraints obtained from submatrices containing two rows corresponding to each of three images and one row from each of the other images are called the quadrilinear constraints. The quadrilinear constraints between image,, and can be written as a product of -, - and -coordinates in the other images and! subdeterminants of (12) Again the first factors consists of projective coordinates, and some combination of these projective coordinates has nonvanishing product, thus the quadrilinear constraints are equivalent to the constraints expressed by subdeterminants from (12). The quadrilinear constraints are usually expressed using the quadrifocal tensor,, with components,, expressed in,,,, and. There turns out to be linearly independent (in ) quadrilinear constraints, see [22], obtained from subdeterminants of (12). This makes it possible to obtain linear constraints on the coefficients of the quadrilinear tensor for each corresponding point in images. Thus having corresponding points in images makes it possible to linearly recover the components of the quadrilinear tensor and from them the camera matrices (up to the previous mentioned ambiguity). Because of the large number of components of the trifocal and quadrifocal tensor ( and respectively) simplifications that reduces these numbers are often needed in calculations. One such simplification is the use of reduced affine coordinates in the images and in the object, see [9]. In this case three corresponding points are chosen as an affine basis in each image and all other coordinates used are affine coordinates with respect to these three basis points. Using this simplification one obtains the reduced fundamental matrix, the reduced trifocal tensor, and the reduced quadrifocal tensor containing,, and components respectively. 4 Dependencies between Multilinear Forms It is obvious that (7) describes all available constraints on the camera matrices in an image sequence. These constraints can be expressed by bilinear, trilinear and quadrilinear constraints. All higher order multilinear constraints are just products of these three different types. Theorem 2. All constraints on the camera matrices in a sequence of images can be expressed in the bilinear, trilinear and quadrilinear constraints between all combinations of, and images respectively. Obviously, this large amount of multilinear constraints is unsatisfactory to work with. In order to reduce the number, assume that the trilinear constraints between every successive triplet of images is known, in the form of the trilinear tensor. Then the each

6 triplet of camera matrices can be calculated up to the projective ambiguity. Then it is a simple task to adapt the whole sequence of camera matrices to each other and obeying all multilinear constraints. Theorem 3. All constraints on the camera matrices in a sequence of images can be expressed by the trilinear constraints between every successive triplet of images. The interesting question now is if we can reduce the representation to bilinear constraints. Introduce the notation (13) Then the three bilinear constraints between images, and can be expressed by # # # (14) with # # # (15) where the notation # " $#! have been used. Thus from the three bilinearities,, and can be recovered up to scale. Since a relative scale between these two vectors can be obtained (if they are not coplanar). Then the relative scale between, and can be calculated and the only ambiguity present in the determination of the camera matrices, and is the projective ambiguity, see also [12]. Note that this relative scale is naturally obtained from the trilinearities in (10) written as (16) This shows that geometrically, the bilinearities between view and and between view and is sufficient to represent the multiple view geometry. Theorem 4. All constraints on the camera matrices in a sequence of images can be expressed geometrically by the bilinear constraints between image and and between image and. Observe that this theorem is true under the hypothesis of general motion, i.e. three successive camera centers are not collinear. Turning to the algebraic point of view, the problem is not so easy. The first difficulty is that the ideal defined by the trilinearities,, is not the same as the ideal defined by the bilinearities,, observed in [4]. For instance for three views three arbitrary points

7 on the trifocal lines (the lines in the images connecting the epipoles from the other two images) in each image obeys all bilinear constraints, but not necessarily all trilinear ones. This corresponds to the fact that trilinearities can distinguish between different points on the trifocal plane, but the bilinearities can not. However, even if the ideals are different it is possible to calculate from in the following algebraic sense, see [10]. Theorem 5. The bilinear ideal is reducible and has a primary decomposition,, where is the ideal corresponding to an arbitrary point on each trifocal line. This theorem, together with the previous one shows that both geometrically and algebraically the bilinearities are sufficient to describe image sequences. However, it is not the best way to treat these problems numerically. 5 Reconstruction using a Subspace Approach In order to reconstruct a scene from an image sequence, one can use the bilinear (or even trilinear) constraints to build up the sequence of camera matrices and the find the reconstruction by triangulation. A numerically better way is to user all available multilinear constraints in (7). Such a reconstruction algorithm will be outlined in this section. 5.1 Notations Introduce the notation " # $ for the coordinates of point number in image number and # # # # (17) # (18) where " # #$ denotes the coordinates of object point number. Describe the depths, in (1), of point in image by the diagonal matrices " # # # # $# # # (19) 5.2 Subspace Formulation Using these notations (4) can be written # # # (20)

8 Denote the linear subspace in spanned by the rows in and by the rows in by and respectively. Then (20) can be interpreted as # (21) where the diagonal matrix is interpreted as an operator that acts on a linear subspace by componentwise multiplication of the diagonal elements. One advantage of this formulation in (21) is that it is independent of the chosen coordinate system in the images, since is the same subspace for every choice of affine coordinates in the images. The advantage of the introduced notations in (17), (18) and (19) is that (1) can be written for images of points:.. (22) For convenience, introduce the notation. Then, multiplying each side with its transpose and dividing with the number of images, (22) can be written (23) The corresponding equation for the subspace analogy follows directly from (21): # (24) where # denotes the sum of the two subspaces and. Our goal now is to use (22), (24) to design an algorithm for calculating, and from the image data. Then gives us the reconstruction (structure, perception) and give us the camera matrices (motion; obtained as the nullspaces of, action). Introduce the matrices # " $ # (25) where is the projection matrix onto. The purpose of the factor in (25) is just to make the size of the entries in more independent on the number of images. Observe that is independent on the chosen basis for since " $ " " $ $ " $ " $ " $ # which implies that is independent on the chosen coordinate system in the image. Now (24) implies that # # (26)

9 since is a dimensional subspace. This can be viewed as another way of describing the multilinear constraints, which is more suitable for numerical calculations. Let, # # be the singular values of and introduce the proximity measure, which measure the degree of -dimensionality of the sum of the subspaces. Then the reconstruction problem can be formulated by a variational formula (27) In the noise free case the minimum value is equal to. When noise is present in the measurements, minimising this variational formula gives a reconstruction that is independent of the chosen coordinate systems and the ordering of the points. This reconstruction can be obtained from the range of, which is equal to. We remark that the proximity measure can not be interpreted in terms of distances in the images, since distances have no meaning in this coordinate independent approach. Instead it is an abstract measure of the dimension of a subspace. 5.3 An Algorithm Reconstruction can be made from an iterative algorithm consisting of the following steps: 1. Start by putting. 2. Calculate from (25). 3. Calculate the singular value decomposition of, i.e. and the proximity measure. If is sufficiently small stop. 4. Let denote the first four rows of, which will be used as an approximation of the object. 5. Use (20) to estimate from and. 6. Goto 2. The criteria on the proximity measure for terminating have to be chosen appropriately. Step 5 above can done giving a result that is independent of both coordinate systems and the ordering of images and points. Let and denote the projection matrices onto form and its orthogonal complement respectively. Then we use (21) in the # (28) where denotes an orthonormal basis for. For details see [11]. We remark that the main difference between this algorithm and the one presented in [23] is that the former one gives a result that is independent of both the ordering of image and points and the choice of coordinate systems in the images, whereas the latter does not. This is due to the use of Euclidean distances between corresponding points and epipolar lines as well as an asymmetry in the use of different images.

10 a b c d e f g h i j k l Figure1. Illustration of the trilinear approach. a)-i) show nine images of a simple scene. In j), some of the extracted points are shown in the affine coordinate system defined by three basis points. In k), the reconstructed camera motion is shown together with a reconstruction of the extracted corner points. l) highlights the reconstructed object. 6 Experiments In this section two different experiments will be presented. Reconstruction using Trilinear Forms We illustrate the trilinear approach with an experiment where 9 images have been taken of a simple scene. These are shown in Fig. 1. Some of the corners are extracted and three corners, one from each box on the floor, are used as an affine basis in each image in order to simplify the algebra (see [9,8]). The corner positions have been used to estimate the trilinear forms and therefrom the reconstruction and the motion of the camera. This motion is presented together with the reconstruction of some of the corner points. Reconstruction using the Subspace Approach Consider the four images of a toy block scene in Figure 2. corresponding points (corners of blocks) were picked out manually and were used as inputs to the algorithm. The results are shown in Figure 2, together with the obtained reconstruction. In order to display the reconstruction a Euclidean representation of all projectively equivalent reconstructions, obtained from the algorithm, has to be chosen. This has been done using a three-dimensional model of the object and selecting the closest possible of the projective reconstructions, in least squares sense. However, this is the only stage where the model has been used.

11 Figure2. Illustration of the performance of the algorithm. Above: Four images of the toy block scene. Below: The logarithm of the proximity measure, the estimated standard deviation and the obtained reconstruction. 7 Conclusions In this paper we have shown that multilinear forms can be used in perception and action tasks in computer vision. Firstly, multilinear forms can be directly used to reconstruct the scene, giving the perception, and to calculate the camera motion, giving a necessary input to action tasks. Secondly, a generic algorithm, based on multilinear forms, for reconstruction and calculation of camera motion has been described. The algorithm is generic in the sense that all corresponding points and all images are treated in an equal way, i.e. no points or images are used as base points or reference image respectively. Furthermore, the algorithm gives a result that is independent on the chosen coordinate systems in the images, which is the natural way to deal with uncalibrated cameras. Although, there is no guarantee that this iterative algorithm will converge, experiments show that the convergence is usually very fast. The algorithms are presented using real images, showing that it is possible to obtain both reconstruction and camera motion. Further investigations would be towards specific action tasks, such as picking up an object or positioning. References 1. Berthilsson, R., Heyden, A., Sparr, G., Recursive Structure and Motion from Image Sequences using Shape and Depth Spaces, to be presented at CVPR 97, Faugeras, O., D., What can be seen in three dimensions with an uncalibrated stereo rig?, ECCV 92, Lecture notes in Computer Science, Vol 588. Ed. G. Sandini, Springer-Verlag, 1992, pp

12 3. Faugeras, O., D., Mourrain, B., On the geometry and algebra on the point and line correspondences between N images, Proc. ICCV 95, IEEE Computer Society Press, 1995, pp Faugeras, O., D., Mourrain, B., About the correspondences of points between N images, Proc. IEEE Workshop on Representation of Visual Scenes, Hartley, R., I., Projective Reconstruction and Invariants from Multiple Images, IEEE Trans. Pattern Anal. Machine Intell., vol. 16, no. 10, pp , Hartley, A linear method for reconstruction from lines and points, Proc. ICCV 95, IEEE Computer Society Press, 1995, pp Heyden, A., Reconstruction and Prediction from Three Images of Uncalibrated Cameras, Proc. 9th Scandinavian Conference on Image Analysis, Ed. Gunilla Borgefors, Uppsala, Sweden, 1995, pp Heyden, A., Reconstruction from Image Sequences by means of Relative Depths, Proc. ICCV 95, IEEE Computer Society Press, 1995, pp An extended version to appear in IJCV, International Journal of Computer Vision, Heyden, A., Åström, K., A Canonical Framework for Sequences of Images, Proc. IEEE Workshop on Representation of Visual Scenes, Heyden, A., Åström, K., Algebraic Varieties in Multiple View Geometry, ECCV 96, Lecture notes in Computer Science, Vol 1065, Ed. B. Buxton, R. Chippola, Springer-Verlag 1996, pp Heyden, A., Projective Structure and Motion from Image Sequences using Subspace Methods, Submitted to SCIA 97, Luong, Q.-T., Vieville, T., Canonic Representations for the Geometries of Multiple Projective Views, ECCV 94, Lecture notes in Computer Science, Vol 800. Ed. Jan-Olof Eklund, Springer-Verlag, 1994, pp Maybank, S., Theory of Reconstruction from Image Motion, Springer-Verlag, Berlin, Heidelberg, New York, Mohr, R., Arbogast, E., It can be done without camera calibration, Pattern Recognition Letters, vol. 12, no. 1, 1991, pp Poelman, C., J., Kanade, T., A Paraperspective Factorization Method for Shape and Motion Recovery, ECCV 94, Lecture notes in Computer Science, Vol 801. Ed. Jan-Olof Eklund, Springer-Verlag, 1994, pp Shashua, A., Trilinearity in Visual Recognition by Alignment, ECCV 94, Lecture notes in Computer Science, Vol 800. Ed. Jan-Olof Eklund, Springer-Verlag, 1994, pp Shashua, A., Werman, M., Trilinearity of Three Perspective Views and its Associated Tensor, Proc. ICCV 95, IEEE Computer Society Press, 1995, pp Sparr, G., An algebraic-analytic method for affine shapes of point configurations, proceedings 7th Scandinavian Conference on Image Analysis, 1991, pp Sparr, G., A Common Framework for Kinetic Depth, Reconstruction and Motion for Deformable Objects, ECCV 94, Lecture notes in Computer Science, Vol 801. Ed. J-O. Eklund, Springer-Verlag 1994, pp Sparr, G., Simultaneous Reconstruction of Scene Structure and Camera Locations from Uncalibrated Image Sequences, proceedings 13th International Conference on Pattern Recognition, 1996, pp Tomasi, C., Kanade, T., Shape and Motion from Image Streams under Orthography: A Factorization Method, IJCV, 9(2): , Triggs, B., Matching Constraints and the Joint Image, Proc. ICCV 95, IEEE Computer Society Press, 1995, pp Sturm, P., Triggs, B., A Factorization Based Algorithm for Multi-Image Projective Structure and Motion, ECCV 96, Lecture notes in Computer Science, Vol Ed. B. Buxton and R. Cipolla, Springer-Verlag 1996, pp

A Canonical Framework for Sequences of Images

A Canonical Framework for Sequences of Images A Canonical Framework for Sequences of Images Anders Heyden, Kalle Åström Dept of Mathematics, Lund University Box 118, S-221 00 Lund, Sweden email: andersp@maths.lth.se kalle@maths.lth.se Abstract This

More information

is used in many dierent applications. We give some examples from robotics. Firstly, a robot equipped with a camera, giving visual information about th

is used in many dierent applications. We give some examples from robotics. Firstly, a robot equipped with a camera, giving visual information about th Geometry and Algebra of Multiple Projective Transformations Anders Heyden Dept of Mathematics, Lund University Box 8, S-22 00 Lund, SWEDEN email: heyden@maths.lth.se Supervisor: Gunnar Sparr Abstract In

More information

Euclidean Reconstruction from Constant Intrinsic Parameters

Euclidean Reconstruction from Constant Intrinsic Parameters uclidean Reconstruction from Constant ntrinsic Parameters nders Heyden, Kalle Åström Dept of Mathematics, Lund University Box 118, S-221 00 Lund, Sweden email: heyden@maths.lth.se, kalle@maths.lth.se bstract

More information

Structure and motion in 3D and 2D from hybrid matching constraints

Structure and motion in 3D and 2D from hybrid matching constraints Structure and motion in 3D and 2D from hybrid matching constraints Anders Heyden, Fredrik Nyberg and Ola Dahl Applied Mathematics Group Malmo University, Sweden {heyden,fredrik.nyberg,ola.dahl}@ts.mah.se

More information

A General Expression of the Fundamental Matrix for Both Perspective and Affine Cameras

A General Expression of the Fundamental Matrix for Both Perspective and Affine Cameras A General Expression of the Fundamental Matrix for Both Perspective and Affine Cameras Zhengyou Zhang* ATR Human Information Processing Res. Lab. 2-2 Hikari-dai, Seika-cho, Soraku-gun Kyoto 619-02 Japan

More information

calibrated coordinates Linear transformation pixel coordinates

calibrated coordinates Linear transformation pixel coordinates 1 calibrated coordinates Linear transformation pixel coordinates 2 Calibration with a rig Uncalibrated epipolar geometry Ambiguities in image formation Stratified reconstruction Autocalibration with partial

More information

Multiple Motion Scene Reconstruction from Uncalibrated Views

Multiple Motion Scene Reconstruction from Uncalibrated Views Multiple Motion Scene Reconstruction from Uncalibrated Views Mei Han C & C Research Laboratories NEC USA, Inc. meihan@ccrl.sj.nec.com Takeo Kanade Robotics Institute Carnegie Mellon University tk@cs.cmu.edu

More information

Unit 3 Multiple View Geometry

Unit 3 Multiple View Geometry Unit 3 Multiple View Geometry Relations between images of a scene Recovering the cameras Recovering the scene structure http://www.robots.ox.ac.uk/~vgg/hzbook/hzbook1.html 3D structure from images Recover

More information

Structure from Motion and Multi- view Geometry. Last lecture

Structure from Motion and Multi- view Geometry. Last lecture Structure from Motion and Multi- view Geometry Topics in Image-Based Modeling and Rendering CSE291 J00 Lecture 5 Last lecture S. J. Gortler, R. Grzeszczuk, R. Szeliski,M. F. Cohen The Lumigraph, SIGGRAPH,

More information

The end of affine cameras

The end of affine cameras The end of affine cameras Affine SFM revisited Epipolar geometry Two-view structure from motion Multi-view structure from motion Planches : http://www.di.ens.fr/~ponce/geomvis/lect3.pptx http://www.di.ens.fr/~ponce/geomvis/lect3.pdf

More information

A Computer Vision Toolbox

A Computer Vision Toolbox A Computer Vision Toolbox Kalle Åström, Anders Heyden, Fredrik Kahl, Rikard Berthilsson, Gunnar Sparr Dept of Mathematics, Lund University Box 118, S-1 Lund, Sweden email: {kalle,heyden,fredrik,rikard,gunnar}@maths.lth.se

More information

Camera Calibration Using Line Correspondences

Camera Calibration Using Line Correspondences Camera Calibration Using Line Correspondences Richard I. Hartley G.E. CRD, Schenectady, NY, 12301. Ph: (518)-387-7333 Fax: (518)-387-6845 Email : hartley@crd.ge.com Abstract In this paper, a method of

More information

Stereo and Epipolar geometry

Stereo and Epipolar geometry Previously Image Primitives (feature points, lines, contours) Today: Stereo and Epipolar geometry How to match primitives between two (multiple) views) Goals: 3D reconstruction, recognition Jana Kosecka

More information

Structure from motion

Structure from motion Structure from motion Structure from motion Given a set of corresponding points in two or more images, compute the camera parameters and the 3D point coordinates?? R 1,t 1 R 2,t 2 R 3,t 3 Camera 1 Camera

More information

Structure from motion

Structure from motion Structure from motion Structure from motion Given a set of corresponding points in two or more images, compute the camera parameters and the 3D point coordinates?? R 1,t 1 R 2,t R 2 3,t 3 Camera 1 Camera

More information

CS 664 Structure and Motion. Daniel Huttenlocher

CS 664 Structure and Motion. Daniel Huttenlocher CS 664 Structure and Motion Daniel Huttenlocher Determining 3D Structure Consider set of 3D points X j seen by set of cameras with projection matrices P i Given only image coordinates x ij of each point

More information

Mathematics of a Multiple Omni-Directional System

Mathematics of a Multiple Omni-Directional System Mathematics of a Multiple Omni-Directional System A. Torii A. Sugimoto A. Imiya, School of Science and National Institute of Institute of Media and Technology, Informatics, Information Technology, Chiba

More information

Stereo Vision. MAN-522 Computer Vision

Stereo Vision. MAN-522 Computer Vision Stereo Vision MAN-522 Computer Vision What is the goal of stereo vision? The recovery of the 3D structure of a scene using two or more images of the 3D scene, each acquired from a different viewpoint in

More information

Structure and Motion for Dynamic Scenes - The Case of Points Moving in Planes

Structure and Motion for Dynamic Scenes - The Case of Points Moving in Planes Structure and Motion for Dynamic Scenes - The Case of Points Moving in Planes Peter Sturm To cite this version: Peter Sturm. Structure and Motion for Dynamic Scenes - The Case of Points Moving in Planes.

More information

Mei Han Takeo Kanade. January Carnegie Mellon University. Pittsburgh, PA Abstract

Mei Han Takeo Kanade. January Carnegie Mellon University. Pittsburgh, PA Abstract Scene Reconstruction from Multiple Uncalibrated Views Mei Han Takeo Kanade January 000 CMU-RI-TR-00-09 The Robotics Institute Carnegie Mellon University Pittsburgh, PA 1513 Abstract We describe a factorization-based

More information

Structure from Motion. Introduction to Computer Vision CSE 152 Lecture 10

Structure from Motion. Introduction to Computer Vision CSE 152 Lecture 10 Structure from Motion CSE 152 Lecture 10 Announcements Homework 3 is due May 9, 11:59 PM Reading: Chapter 8: Structure from Motion Optional: Multiple View Geometry in Computer Vision, 2nd edition, Hartley

More information

Multiple View Geometry in Computer Vision Second Edition

Multiple View Geometry in Computer Vision Second Edition Multiple View Geometry in Computer Vision Second Edition Richard Hartley Australian National University, Canberra, Australia Andrew Zisserman University of Oxford, UK CAMBRIDGE UNIVERSITY PRESS Contents

More information

Flexible 3D Models from Uncalibrated Cameras

Flexible 3D Models from Uncalibrated Cameras Flexible 3D Models from Uncalibrated Cameras T.F.Cootes, E.C. Di Mauro, C.J.Taylor, AXanitis Department of Medical Biophysics, University of Manchester Manchester M13 9PT email: bim@svl.smb.man.ac.uk Abstract

More information

Linear Multi View Reconstruction and Camera Recovery Using a Reference Plane

Linear Multi View Reconstruction and Camera Recovery Using a Reference Plane International Journal of Computer Vision 49(2/3), 117 141, 2002 c 2002 Kluwer Academic Publishers. Manufactured in The Netherlands. Linear Multi View Reconstruction and Camera Recovery Using a Reference

More information

Two-View Geometry (Course 23, Lecture D)

Two-View Geometry (Course 23, Lecture D) Two-View Geometry (Course 23, Lecture D) Jana Kosecka Department of Computer Science George Mason University http://www.cs.gmu.edu/~kosecka General Formulation Given two views of the scene recover the

More information

Partial Calibration and Mirror Shape Recovery for Non-Central Catadioptric Systems

Partial Calibration and Mirror Shape Recovery for Non-Central Catadioptric Systems Partial Calibration and Mirror Shape Recovery for Non-Central Catadioptric Systems Abstract In this paper we present a method for mirror shape recovery and partial calibration for non-central catadioptric

More information

Combining Two-view Constraints For Motion Estimation

Combining Two-view Constraints For Motion Estimation ombining Two-view onstraints For Motion Estimation Venu Madhav Govindu Somewhere in India venu@narmada.org Abstract In this paper we describe two methods for estimating the motion parameters of an image

More information

Creating 3D Models with Uncalibrated Cameras

Creating 3D Models with Uncalibrated Cameras Creating D Models with Uncalibrated Cameras Mei Han Takeo Kanade Robotics Institute, Carnegie Mellon University meihan,tk@cs.cmu.edu Abstract We describe a factorization-based method to recover D models

More information

Minimal Projective Reconstruction for Combinations of Points and Lines in Three Views

Minimal Projective Reconstruction for Combinations of Points and Lines in Three Views Minimal Projective Reconstruction for Combinations of Points and Lines in Three Views Magnus Oskarsson, Andrew Zisserman and Kalle Åström Centre for Mathematical Sciences Lund University,SE 221 00 Lund,

More information

Mei Han Takeo Kanade. August Carnegie Mellon University. Pittsburgh, PA 15213

Mei Han Takeo Kanade. August Carnegie Mellon University. Pittsburgh, PA 15213 Perspective Factorization Methods for Euclidean Reconstruction Mei Han Takeo Kanade August 1999 CMU-RI-TR-99-22 The Robotics Institute Carnegie Mellon University Pittsburgh, PA 15213 @1999 Carnegie Mellon

More information

Lecture 6 Stereo Systems Multi- view geometry Professor Silvio Savarese Computational Vision and Geometry Lab Silvio Savarese Lecture 6-24-Jan-15

Lecture 6 Stereo Systems Multi- view geometry Professor Silvio Savarese Computational Vision and Geometry Lab Silvio Savarese Lecture 6-24-Jan-15 Lecture 6 Stereo Systems Multi- view geometry Professor Silvio Savarese Computational Vision and Geometry Lab Silvio Savarese Lecture 6-24-Jan-15 Lecture 6 Stereo Systems Multi- view geometry Stereo systems

More information

Multiple View Geometry of Projector-Camera Systems from Virtual Mutual Projection

Multiple View Geometry of Projector-Camera Systems from Virtual Mutual Projection Multiple View Geometry of rojector-camera Systems from Virtual Mutual rojection Shuhei Kobayashi, Fumihiko Sakaue, and Jun Sato Department of Computer Science and Engineering Nagoya Institute of Technology

More information

Announcements. Stereo

Announcements. Stereo Announcements Stereo Homework 2 is due today, 11:59 PM Homework 3 will be assigned today Reading: Chapter 7: Stereopsis CSE 152 Lecture 8 Binocular Stereopsis: Mars Given two images of a scene where relative

More information

A Factorization Method for Structure from Planar Motion

A Factorization Method for Structure from Planar Motion A Factorization Method for Structure from Planar Motion Jian Li and Rama Chellappa Center for Automation Research (CfAR) and Department of Electrical and Computer Engineering University of Maryland, College

More information

Epipolar geometry. x x

Epipolar geometry. x x Two-view geometry Epipolar geometry X x x Baseline line connecting the two camera centers Epipolar Plane plane containing baseline (1D family) Epipoles = intersections of baseline with image planes = projections

More information

Rectification and Distortion Correction

Rectification and Distortion Correction Rectification and Distortion Correction Hagen Spies March 12, 2003 Computer Vision Laboratory Department of Electrical Engineering Linköping University, Sweden Contents Distortion Correction Rectification

More information

Research Article Projective Invariants from Multiple Images: A Direct and Linear Method

Research Article Projective Invariants from Multiple Images: A Direct and Linear Method Mathematical Problems in Engineering Volume 26, Article ID 852364, 4 pages http://dx.doi.org/.55/26/852364 Research Article Projective Invariants from Multiple Images: A Direct and Linear Method Yuanbin

More information

CS231M Mobile Computer Vision Structure from motion

CS231M Mobile Computer Vision Structure from motion CS231M Mobile Computer Vision Structure from motion - Cameras - Epipolar geometry - Structure from motion Pinhole camera Pinhole perspective projection f o f = focal length o = center of the camera z y

More information

Lecture 9: Epipolar Geometry

Lecture 9: Epipolar Geometry Lecture 9: Epipolar Geometry Professor Fei Fei Li Stanford Vision Lab 1 What we will learn today? Why is stereo useful? Epipolar constraints Essential and fundamental matrix Estimating F (Problem Set 2

More information

Integrating Local Affine into Global Projective Images in the Joint Image Space

Integrating Local Affine into Global Projective Images in the Joint Image Space Integrating Local Affine into Global Projective Images in the Joint Image Space P. Anandan and Shai Avidan Microsoft Research, One Microsoft Way, Redmond, WA 98052, USA, {anandan,avidan}@microsoft.com

More information

Euclidean Reconstruction and Auto-Calibration from Continuous Motion

Euclidean Reconstruction and Auto-Calibration from Continuous Motion Euclidean Reconstruction and Auto-Calibration from Continuous Motion Fredrik Kahl and Anders Heyden Λ Centre for Mathematical Sciences Lund University Box 8, SE- Lund, Sweden {fredrik, andersp}@maths.lth.se

More information

Euclidean Reconstruction Independent on Camera Intrinsic Parameters

Euclidean Reconstruction Independent on Camera Intrinsic Parameters Euclidean Reconstruction Independent on Camera Intrinsic Parameters Ezio MALIS I.N.R.I.A. Sophia-Antipolis, FRANCE Adrien BARTOLI INRIA Rhone-Alpes, FRANCE Abstract bundle adjustment techniques for Euclidean

More information

Multiple-View Structure and Motion From Line Correspondences

Multiple-View Structure and Motion From Line Correspondences ICCV 03 IN PROCEEDINGS OF THE IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION, NICE, FRANCE, OCTOBER 003. Multiple-View Structure and Motion From Line Correspondences Adrien Bartoli Peter Sturm INRIA

More information

Visual Recognition: Image Formation

Visual Recognition: Image Formation Visual Recognition: Image Formation Raquel Urtasun TTI Chicago Jan 5, 2012 Raquel Urtasun (TTI-C) Visual Recognition Jan 5, 2012 1 / 61 Today s lecture... Fundamentals of image formation You should know

More information

Machine vision. Summary # 11: Stereo vision and epipolar geometry. u l = λx. v l = λy

Machine vision. Summary # 11: Stereo vision and epipolar geometry. u l = λx. v l = λy 1 Machine vision Summary # 11: Stereo vision and epipolar geometry STEREO VISION The goal of stereo vision is to use two cameras to capture 3D scenes. There are two important problems in stereo vision:

More information

EXPERIMENTAL RESULTS ON THE DETERMINATION OF THE TRIFOCAL TENSOR USING NEARLY COPLANAR POINT CORRESPONDENCES

EXPERIMENTAL RESULTS ON THE DETERMINATION OF THE TRIFOCAL TENSOR USING NEARLY COPLANAR POINT CORRESPONDENCES EXPERIMENTAL RESULTS ON THE DETERMINATION OF THE TRIFOCAL TENSOR USING NEARLY COPLANAR POINT CORRESPONDENCES Camillo RESSL Institute of Photogrammetry and Remote Sensing University of Technology, Vienna,

More information

Automatic Line Matching across Views

Automatic Line Matching across Views Automatic Line Matching across Views Cordelia Schmid and Andrew Zisserman Department of Engineering Science, University of Oxford Parks Road, Oxford, UK OX1 3PJ Abstract This paper presents a new method

More information

Factorization with Missing and Noisy Data

Factorization with Missing and Noisy Data Factorization with Missing and Noisy Data Carme Julià, Angel Sappa, Felipe Lumbreras, Joan Serrat, and Antonio López Computer Vision Center and Computer Science Department, Universitat Autònoma de Barcelona,

More information

BIL Computer Vision Apr 16, 2014

BIL Computer Vision Apr 16, 2014 BIL 719 - Computer Vision Apr 16, 2014 Binocular Stereo (cont d.), Structure from Motion Aykut Erdem Dept. of Computer Engineering Hacettepe University Slide credit: S. Lazebnik Basic stereo matching algorithm

More information

Hand-Eye Calibration from Image Derivatives

Hand-Eye Calibration from Image Derivatives Hand-Eye Calibration from Image Derivatives Abstract In this paper it is shown how to perform hand-eye calibration using only the normal flow field and knowledge about the motion of the hand. The proposed

More information

Multiple View Geometry

Multiple View Geometry Multiple View Geometry CS 6320, Spring 2013 Guest Lecture Marcel Prastawa adapted from Pollefeys, Shah, and Zisserman Single view computer vision Projective actions of cameras Camera callibration Photometric

More information

Rigid Multiview Varieties

Rigid Multiview Varieties Joe Kileel University of California, Berkeley January 9, 2016 Nonlinear Algebra JMM, Seattle Joe Kileel Preprint arxiv:1509.03257 Michael Joswig Bernd Sturmfels André Wagner Algebraic vision Multiview

More information

3D Geometry and Camera Calibration

3D Geometry and Camera Calibration 3D Geometry and Camera Calibration 3D Coordinate Systems Right-handed vs. left-handed x x y z z y 2D Coordinate Systems 3D Geometry Basics y axis up vs. y axis down Origin at center vs. corner Will often

More information

Announcements. Stereo

Announcements. Stereo Announcements Stereo Homework 1 is due today, 11:59 PM Homework 2 will be assigned on Thursday Reading: Chapter 7: Stereopsis CSE 252A Lecture 8 Binocular Stereopsis: Mars Given two images of a scene where

More information

Plane + Parallax, Tensors and Factorization

Plane + Parallax, Tensors and Factorization Plane + Parallax, Tensors and Factorization Bill Triggs INRIA Rhône-Alpes, 655 avenue de l Europe, 38330 Montbonnot, France. Bill.Triggs@inrialpes.fr http://www.inrialpes.fr/movi/people/ Triggs Abstract

More information

Index. 3D reconstruction, point algorithm, point algorithm, point algorithm, point algorithm, 263

Index. 3D reconstruction, point algorithm, point algorithm, point algorithm, point algorithm, 263 Index 3D reconstruction, 125 5+1-point algorithm, 284 5-point algorithm, 270 7-point algorithm, 265 8-point algorithm, 263 affine point, 45 affine transformation, 57 affine transformation group, 57 affine

More information

C280, Computer Vision

C280, Computer Vision C280, Computer Vision Prof. Trevor Darrell trevor@eecs.berkeley.edu Lecture 11: Structure from Motion Roadmap Previous: Image formation, filtering, local features, (Texture) Tues: Feature-based Alignment

More information

Index. 3D reconstruction, point algorithm, point algorithm, point algorithm, point algorithm, 253

Index. 3D reconstruction, point algorithm, point algorithm, point algorithm, point algorithm, 253 Index 3D reconstruction, 123 5+1-point algorithm, 274 5-point algorithm, 260 7-point algorithm, 255 8-point algorithm, 253 affine point, 43 affine transformation, 55 affine transformation group, 55 affine

More information

Image Transfer Methods. Satya Prakash Mallick Jan 28 th, 2003

Image Transfer Methods. Satya Prakash Mallick Jan 28 th, 2003 Image Transfer Methods Satya Prakash Mallick Jan 28 th, 2003 Objective Given two or more images of the same scene, the objective is to synthesize a novel view of the scene from a view point where there

More information

Computer Vision Lecture 17

Computer Vision Lecture 17 Computer Vision Lecture 17 Epipolar Geometry & Stereo Basics 13.01.2015 Bastian Leibe RWTH Aachen http://www.vision.rwth-aachen.de leibe@vision.rwth-aachen.de Announcements Seminar in the summer semester

More information

Elements of Computer Vision: Multiple View Geometry. 1 Introduction. 2 Elements of Geometry. Andrea Fusiello

Elements of Computer Vision: Multiple View Geometry. 1 Introduction. 2 Elements of Geometry. Andrea Fusiello Elements of Computer Vision: Multiple View Geometry. Andrea Fusiello http://www.sci.univr.it/~fusiello July 11, 2005 c Copyright by Andrea Fusiello. This work is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike

More information

Computer Vision Lecture 17

Computer Vision Lecture 17 Announcements Computer Vision Lecture 17 Epipolar Geometry & Stereo Basics Seminar in the summer semester Current Topics in Computer Vision and Machine Learning Block seminar, presentations in 1 st week

More information

Structure from Motion

Structure from Motion Structure from Motion Outline Bundle Adjustment Ambguities in Reconstruction Affine Factorization Extensions Structure from motion Recover both 3D scene geoemetry and camera positions SLAM: Simultaneous

More information

Stereo CSE 576. Ali Farhadi. Several slides from Larry Zitnick and Steve Seitz

Stereo CSE 576. Ali Farhadi. Several slides from Larry Zitnick and Steve Seitz Stereo CSE 576 Ali Farhadi Several slides from Larry Zitnick and Steve Seitz Why do we perceive depth? What do humans use as depth cues? Motion Convergence When watching an object close to us, our eyes

More information

Factorization Methods for Projective Structure and Motion

Factorization Methods for Projective Structure and Motion Factorization Methods for Projective Structure and Motion Bill Triggs LIFIA, INRIA Rhône-Alpes, 6, avenue Félix Viallet, 383 Grenoble, France. Bill.Triggs@imag.fr Abstract This paper describes a family

More information

CEE598 - Visual Sensing for Civil Infrastructure Eng. & Mgmt.

CEE598 - Visual Sensing for Civil Infrastructure Eng. & Mgmt. CEE598 - Visual Sensing for Civil Infrastructure Eng. & Mgmt. Session 4 Affine Structure from Motion Mani Golparvar-Fard Department of Civil and Environmental Engineering 329D, Newmark Civil Engineering

More information

A Summary of Projective Geometry

A Summary of Projective Geometry A Summary of Projective Geometry Copyright 22 Acuity Technologies Inc. In the last years a unified approach to creating D models from multiple images has been developed by Beardsley[],Hartley[4,5,9],Torr[,6]

More information

Feature Transfer and Matching in Disparate Stereo Views through the use of Plane Homographies

Feature Transfer and Matching in Disparate Stereo Views through the use of Plane Homographies Feature Transfer and Matching in Disparate Stereo Views through the use of Plane Homographies M. Lourakis, S. Tzurbakis, A. Argyros, S. Orphanoudakis Computer Vision and Robotics Lab (CVRL) Institute of

More information

Week 2: Two-View Geometry. Padua Summer 08 Frank Dellaert

Week 2: Two-View Geometry. Padua Summer 08 Frank Dellaert Week 2: Two-View Geometry Padua Summer 08 Frank Dellaert Mosaicking Outline 2D Transformation Hierarchy RANSAC Triangulation of 3D Points Cameras Triangulation via SVD Automatic Correspondence Essential

More information

The Geometry of Dynamic Scenes On Coplanar and Convergent Linear Motions Embedded in 3D Static Scenes

The Geometry of Dynamic Scenes On Coplanar and Convergent Linear Motions Embedded in 3D Static Scenes EXTENDED VERSION SHORT VERSION APPEARED IN THE 13TH BMVC, CARDIFF, SEPTEMBER 2002. The Geometry of Dynamic Scenes On Coplanar and Convergent Linear Motions Embedded in 3D Static Scenes Adrien Bartoli,

More information

Camera Calibration from the Quasi-affine Invariance of Two Parallel Circles

Camera Calibration from the Quasi-affine Invariance of Two Parallel Circles Camera Calibration from the Quasi-affine Invariance of Two Parallel Circles Yihong Wu, Haijiang Zhu, Zhanyi Hu, and Fuchao Wu National Laboratory of Pattern Recognition, Institute of Automation, Chinese

More information

Two-view geometry Computer Vision Spring 2018, Lecture 10

Two-view geometry Computer Vision Spring 2018, Lecture 10 Two-view geometry http://www.cs.cmu.edu/~16385/ 16-385 Computer Vision Spring 2018, Lecture 10 Course announcements Homework 2 is due on February 23 rd. - Any questions about the homework? - How many of

More information

Projective reconstruction from line-correspondences in multiple uncalibrated images

Projective reconstruction from line-correspondences in multiple uncalibrated images Pattern Recognition 39 (2006) 889 896 www.elsevier.com/locate/patcog Projective reconstruction from line-correspondences in multiple uncalibrated images A.W.K. Tang, T.P. Ng, Y.S. Hung, C.H. Leung Department

More information

3D Model Acquisition by Tracking 2D Wireframes

3D Model Acquisition by Tracking 2D Wireframes 3D Model Acquisition by Tracking 2D Wireframes M. Brown, T. Drummond and R. Cipolla {96mab twd20 cipolla}@eng.cam.ac.uk Department of Engineering University of Cambridge Cambridge CB2 1PZ, UK Abstract

More information

Motion Segmentation Based on Factorization Method and Discriminant Criterion

Motion Segmentation Based on Factorization Method and Discriminant Criterion Motion Segmentation Based on Factorization Method and Discriminant Criterion Naoyuki ICHIMURA Electrotechnical Laboratory 1-1-, Umezono, Tsukuba Ibaraki, -6 Japan ichimura@etl.go.jp Abstract A motion segmentation

More information

Synchronized Ego-Motion Recovery of Two Face-to-Face Cameras

Synchronized Ego-Motion Recovery of Two Face-to-Face Cameras Synchronized Ego-Motion Recovery of Two Face-to-Face Cameras Jinshi Cui, Yasushi Yagi, Hongbin Zha, Yasuhiro Mukaigawa, and Kazuaki Kondo State Key Lab on Machine Perception, Peking University, China {cjs,zha}@cis.pku.edu.cn

More information

Coplanar circles, quasi-affine invariance and calibration

Coplanar circles, quasi-affine invariance and calibration Image and Vision Computing 24 (2006) 319 326 www.elsevier.com/locate/imavis Coplanar circles, quasi-affine invariance and calibration Yihong Wu *, Xinju Li, Fuchao Wu, Zhanyi Hu National Laboratory of

More information

MULTIPLE VIEW GEOMETRY

MULTIPLE VIEW GEOMETRY Chapter 3 MULTIPLE VIEW GEOMETRY Anders Heyden and Marc Pollefeys 3.1 Introduction There exist intricate geometric relations between multiple views of a 3D scene. These relations are related to the camera

More information

But First: Multi-View Projective Geometry

But First: Multi-View Projective Geometry View Morphing (Seitz & Dyer, SIGGRAPH 96) Virtual Camera Photograph Morphed View View interpolation (ala McMillan) but no depth no camera information Photograph But First: Multi-View Projective Geometry

More information

Epipolar Geometry in Stereo, Motion and Object Recognition

Epipolar Geometry in Stereo, Motion and Object Recognition Epipolar Geometry in Stereo, Motion and Object Recognition A Unified Approach by GangXu Department of Computer Science, Ritsumeikan University, Kusatsu, Japan and Zhengyou Zhang INRIA Sophia-Antipolis,

More information

Factorization Method Using Interpolated Feature Tracking via Projective Geometry

Factorization Method Using Interpolated Feature Tracking via Projective Geometry Factorization Method Using Interpolated Feature Tracking via Projective Geometry Hideo Saito, Shigeharu Kamijima Department of Information and Computer Science, Keio University Yokohama-City, 223-8522,

More information

A Stratified Approach to Metric Self-Calibration

A Stratified Approach to Metric Self-Calibration A Stratified Approach to Metric Self-Calibration Marc Pollefeys and Luc Van Gool K.U.Leuven-MI2 Belgium firstname.lastname@esat.kuleuven.ac.be Abstract Camera calibration is essential to many computer

More information

Partial Calibration and Mirror Shape Recovery for Non-Central Catadioptric Systems

Partial Calibration and Mirror Shape Recovery for Non-Central Catadioptric Systems Partial Calibration and Mirror Shape Recovery for Non-Central Catadioptric Systems Nuno Gonçalves and Helder Araújo Institute of Systems and Robotics - Coimbra University of Coimbra Polo II - Pinhal de

More information

A Comparison of Projective Reconstruction Methods for Pairs of Views

A Comparison of Projective Reconstruction Methods for Pairs of Views A Comparison of Projective Reconstruction Methods for Pairs of Views Rothwell, C, Csurka, G and Faugeras, O INRIA, Sophia Antipolis, France email: [crothwel,csurka,faugeras]@sophiainriafr Abstract Recently,

More information

55:148 Digital Image Processing Chapter 11 3D Vision, Geometry

55:148 Digital Image Processing Chapter 11 3D Vision, Geometry 55:148 Digital Image Processing Chapter 11 3D Vision, Geometry Topics: Basics of projective geometry Points and hyperplanes in projective space Homography Estimating homography from point correspondence

More information

Today. Stereo (two view) reconstruction. Multiview geometry. Today. Multiview geometry. Computational Photography

Today. Stereo (two view) reconstruction. Multiview geometry. Today. Multiview geometry. Computational Photography Computational Photography Matthias Zwicker University of Bern Fall 2009 Today From 2D to 3D using multiple views Introduction Geometry of two views Stereo matching Other applications Multiview geometry

More information

Contents. 1 Introduction Background Organization Features... 7

Contents. 1 Introduction Background Organization Features... 7 Contents 1 Introduction... 1 1.1 Background.... 1 1.2 Organization... 2 1.3 Features... 7 Part I Fundamental Algorithms for Computer Vision 2 Ellipse Fitting... 11 2.1 Representation of Ellipses.... 11

More information

Fundamental Matrix & Structure from Motion

Fundamental Matrix & Structure from Motion Fundamental Matrix & Structure from Motion Instructor - Simon Lucey 16-423 - Designing Computer Vision Apps Today Transformations between images Structure from Motion The Essential Matrix The Fundamental

More information

Vision Review: Image Formation. Course web page:

Vision Review: Image Formation. Course web page: Vision Review: Image Formation Course web page: www.cis.udel.edu/~cer/arv September 10, 2002 Announcements Lecture on Thursday will be about Matlab; next Tuesday will be Image Processing The dates some

More information

Structure from Motion. Prof. Marco Marcon

Structure from Motion. Prof. Marco Marcon Structure from Motion Prof. Marco Marcon Summing-up 2 Stereo is the most powerful clue for determining the structure of a scene Another important clue is the relative motion between the scene and (mono)

More information

The 3D Line Motion Matrix and Alignment of Line Reconstructions

The 3D Line Motion Matrix and Alignment of Line Reconstructions The 3D Line Motion Matrix and Alignment of Line Reconstructions Adrien Bartoli, Peter Sturm To cite this version: Adrien Bartoli, Peter Sturm. The 3D Line Motion Matrix and Alignment of Line Reconstructions.

More information

Lecture 10: Multi-view geometry

Lecture 10: Multi-view geometry Lecture 10: Multi-view geometry Professor Stanford Vision Lab 1 What we will learn today? Review for stereo vision Correspondence problem (Problem Set 2 (Q3)) Active stereo vision systems Structure from

More information

An idea which can be used once is a trick. If it can be used more than once it becomes a method

An idea which can be used once is a trick. If it can be used more than once it becomes a method An idea which can be used once is a trick. If it can be used more than once it becomes a method - George Polya and Gabor Szego University of Texas at Arlington Rigid Body Transformations & Generalized

More information

Epipolar Geometry and the Essential Matrix

Epipolar Geometry and the Essential Matrix Epipolar Geometry and the Essential Matrix Carlo Tomasi The epipolar geometry of a pair of cameras expresses the fundamental relationship between any two corresponding points in the two image planes, and

More information

Motion Segmentation Based on Factorization Method and Discriminant Criterion

Motion Segmentation Based on Factorization Method and Discriminant Criterion Motion Segmentation Based on Factorization Method and Discriminant Criterion Naoyuki ICHIMURA Electrotechnical Laboratory 1-1-, Umezono, Tsukuba Ibaraki, 35-8568 Japan ichimura@etl.go.jp Abstract A motion

More information

CS231A Course Notes 4: Stereo Systems and Structure from Motion

CS231A Course Notes 4: Stereo Systems and Structure from Motion CS231A Course Notes 4: Stereo Systems and Structure from Motion Kenji Hata and Silvio Savarese 1 Introduction In the previous notes, we covered how adding additional viewpoints of a scene can greatly enhance

More information

Recovering structure from a single view Pinhole perspective projection

Recovering structure from a single view Pinhole perspective projection EPIPOLAR GEOMETRY The slides are from several sources through James Hays (Brown); Silvio Savarese (U. of Michigan); Svetlana Lazebnik (U. Illinois); Bill Freeman and Antonio Torralba (MIT), including their

More information

Rectification for Any Epipolar Geometry

Rectification for Any Epipolar Geometry Rectification for Any Epipolar Geometry Daniel Oram Advanced Interfaces Group Department of Computer Science University of Manchester Mancester, M13, UK oramd@cs.man.ac.uk Abstract This paper proposes

More information

Plane-based Calibration Algorithm for Multi-camera Systems via Factorization of Homography Matrices

Plane-based Calibration Algorithm for Multi-camera Systems via Factorization of Homography Matrices Plane-based Calibration Algorithm for Multi-camera Systems via Factorization of Homography Matrices Toshio Ueshiba Fumiaki Tomita National Institute of Advanced Industrial Science and Technology (AIST)

More information

Lecture 6 Stereo Systems Multi-view geometry

Lecture 6 Stereo Systems Multi-view geometry Lecture 6 Stereo Systems Multi-view geometry Professor Silvio Savarese Computational Vision and Geometry Lab Silvio Savarese Lecture 6-5-Feb-4 Lecture 6 Stereo Systems Multi-view geometry Stereo systems

More information