The appearance of non-rigid objects detected and tracked in video streams is highly variable and therefore makes the identification of similar objects very complex. Furthermore, indexing and searching of them represent a very challenging problem in computer vision. The paper presents a framework for object-based matching that increases the robustness of existing feature detectors used for object recognition. The Gaussian mixture densities are used to model intra-shot variations of observed features of tracked objects. This process is achieved by the expectation maximization algorithm which separates feature distributions given by a tracked object into homogeneous clusters. We use seven different variants of Gaussian mixtures and the Bayes information criterion to identify the best structure of the data (model and parameters). Experiments are conducted on a video sequence of fifteen different tracked objects and comparison in the performance of the mixture approach and the two key-frame methods is analyzed and reported.

Many computer vision tasks rely on feature extraction. Interest points are such features. This paper shows that interest points are geometrically stable under different transformations and have high information content (distinctiveness). These two properties make interest points very successful in the contest of image matching. To measure these two properties quantitatively, we introduce two evaluation criteria: repeatability rate and information content. The quality of the interest points depends on the detector used. In this paper several detectors are compared according to the criteria specified above. We determine which detector gives the best results and show that it satisfies the criteria well.

Concerns the automation of the camera network design process in order to obtain accurate 3D measurements. We restrict ourselves to the problem where the camera positions are limited only by the incidence angle constraint and it is simplified to the case where the cameras remain at a fixed distance to the set of target points to be measured. The main question addressed is where to place the cameras in order to obtain the minimal 3D error. From this question several subproblems arise: how to develop a good criterion to judge the configuration; what conditions are needed for the system to work; which are the interrelated aspects involved in the development of the system; and how to optimize the placement of the camera. From these initial questions the choice of a criterion combined with an optimization process is the key concept. The approach can be divided into two main components. Firstly, we develop an uncertainty analysis based on error propagation. This allows us to express an error criterion to be minimized. Secondly, we present an evolutionary optimization method similar to genetic algorithm, which optimizes this criterion.

This paper presents the first version of our platform for automatically building the structure of a video sequence. The first application uses semi-automatic tools based only on image analysis for building interactive videos: decomposing the video into shots, extracting and tracking objects within each shot and linking occurrences of similar objects among the shots. The second application provides the end user with a powerful browser to navigate through any preprocessed hypervideo.

This paper addresses the problem of computing the fundamental matrix which describes a geometric relationship between a pair of stereo images: the epipolar geometry. In the uncalibrated case, epipolar geometry captures all the 3D information available from the scene. It is of central importance for problems such as 3D reconstruction, self-calibration and feature tracking. Hence, the computation of the fundamental matrix is of great interest. The existing classical methods14 use two steps: a linear step followed by a nonlinear one. However, in some cases, the linear step does not yield a close form solution for the fundamental matrix, resulting in more iterations for the nonlinear step which is not guaranteed to converge to the correct solution. In this paper, a novel method based on virtual parallax is proposed. The problem is formulated differently; instead of computing directly the 3 × 3 fundamental matrix, we compute a homography with one epipole position, and show that this is equivalent to computing the fundamental matrix. Simple equations are derived by reducing the number of parameters to estimate. As a consequence, we obtain an accurate fundamental matrix with a stable linear computation. Experiments with simulated and real images validate our method and clearly show the improvement over the classical 8-point method.

We are addressing the problem of Euclidean reconstruction with an uncalibrated affine camera and the calibration of this camera. We investigate constraints under which the Euclidean shape and motion problem becomes linear. The theoretical study described in this paper leads us to impose some practical constraints that the camera is mounted onto a robot arm and that the robot is executing controlled motions whose parameters are known. The affine camera model considered here is just an approximation of the true projective mapping. Nevertheless, there is a large number of applications for which the camera is allowed to be at some distance from the scene and under these circumstances the affine model is a good approximation. The fact that we deal with an uncalibrated camera is an advantage over previous methods because we do not rely any more on the tedious task of camera calibration. The experimental results obtained show that the method described compares favourably with other similar methods.

This paper addresses the problem of retrieving images from large image databases. The method is based on local grayvalue invariants which are computed at automatically detected interest points. A voting algorithm and semilocal constraints make retrieval possible. Indexing allows for efficient retrieval from a database of more than 1,000 images. Experimental results show correct retrieval in the case of partial visibility, similarity transformations, extraneous features, and small perspective deformations.

This paper addresses the problem of recognizing objects in large image databases. The method is based on local characteristics which are invariant to similarity transformations in the image. These characteristics are computed at automatically detected keypoints using the greyvalue signal. The method therefore works on images such as paintings for which geometry based recognition fails. Due to the locality of the method, images can be recognized being given part of an image and in the presence of occlusions. Applying a voting algorithm and semi-local constraints makes the method robust to noise, scene clutter and small perspective deformations. Experiments show an efficient recognition for different types of images. The approach has been validated on an image database containing 1020 images, some of them being very similar by structure, texture or shape.

The paper presents a general method to retrieve images from large databases using images as queries. The method is based on local characteristics which are robust to the group of similarity transformations in the image. Images can be retrieved even if they are translated, rotated or scaled. Due to the locality of the characterization, images can be retrieved even if only a small part of the image is given as well as in the presence of occlusions. A voting algorithm, following the idea of a Hough transform, and semi local constraints allow us to develop a new method which is robust to noise, to scene clutter and small perspective deformations. Experiments show an efficient recognition for different types of images. The approach has been validated on an image database containing 1020 images, some of them being very similar by structure, texture or shape.

We estimate the internal and external parameters of the camera, and simultaneously the distortion's parameters. Our aim is the selection of the best distortion model, using statistical test for the importance of distortion parameters. Also we examine the accuracy in camera parameter estimation and 3D reconstruction, in relation with the noise in the image. We answer the question "until which level of noise, is it possible to obtain a good camera parameter estimation, and from there a good reconstruction?". Experiments are evaluated on simulated data and a rest is performed with real data.

International audience

The paper addresses the problem of computing the fundamental matrix which describes a geometric relationship between a pair of stereo images: the epipolar geometry. We propose a novel method based on virtual parallax. Instead of computing directly the 3/spl times/3 fundamental matrix, we compute a homography with one epipole position, and show that this is equivalent to computing the fundamental matrix. Simple equations are derived by reducing the number of parameters to estimate. As a consequence, we obtain an accurate fundamental matrix of rank two with a stable linear computation. Experiments with simulated and real images validate our method and clearly show the improvement over existing methods.>

This paper presents a method for doing motion segmentation for autonomous vehicles which drive on planar surfaces. There are two distinct types of independent motion that may occur within an image sequence taken from a moving vehicle. The first generic type of independent motion is when the projected motion of points on the independent object violate the epipolar constraint. The second case is where the epipolar constraint is not violated. This paper demonstrates that it is possible to detect this second type of independent motion by looking for progressive dis-occlusion of the road. A novel collision prediction method is also given. The method predicts the projection of a corridor down which the AGV will travel. This prediction may be used for time to contact collision prediction and the corridor width embodies an estimate of the vehicles size.>

This paper considers the problem of merging a set of distinct three dimensional B-spline surface patches, which are reconstructed from observations of the motion of occluding contours in image sequences. We propose an original method of fusing these partially overlapping patches in order to obtain a whole surface. This approach is based on a triangular mesh and surface interpolation through regularized uniform bicubic B-spline surface patches. Experimental results are presented for both synthetic and real data.

It is shown how relative 3-D reconstruction for point correspondence of multiple images from uncalibrated cameras can be achieved through reference points. The original contributions with respect to other related works in the field are a direct global method for relative 3-D reconstruction and a geometrical method to select a correct set of reference points among all image points. Experimental results from both simulated and real image sequences are presented.>

It is possible to recover the three-dimensional structure of a scene using images taken with uncalibrated cameras and pixel correspondences betweeen these images. But such reconstruction can only be performed up to a projective transformation of the 3-D space. Therefore, constraints have to be put on the reconstructed data to get the reconstruction in the Euclidean space. Such constraints arise from knowledge of the scene, such as the location of points, geometrical constraints on lines, etc. The kind of constraints that have to be added are discussed, and it is shown how they can be fed in a general framework. Experimental results on real data prove the feasibility, and experiments on simulated data address the accuracy of the results.>

A method for calibrating single scanline CCD cameras is described. It is shown that the more classical 2D camera calibration techniques are necessary but not sufficient for solving the 1D camera calibration problem. A model for single scanline cameras is proposed, and a two-step procedure for estimating its parameters is provided. It is also shown how the extrinsic camera parameters can be determined geometrically without making explicit the intrinsic camera parameters. The accuracy of the calibration method is analyzed through an application example.>

The observation of curved contours in image sequences is used in egomotion estimation and in surface reconstruction. An egomotion technique that can be applied when no point or straight line correspondences are available is presented. It generalizes egomotion to the case of arbitrarily shaped contours, which is especially valuable in the case of nonpolyhedral objects. The computation uses a very simple finite differences scheme and quickly provides a good estimation of the motion parameters. Experiments conducted on synthetic and real data show the validity of the approach.>

A method for calibrating linear cameras is described. It is shown that the more classical 2-D camera calibration techniques are necessary but not sufficient for solving the linear-camera calibration problem. A model for linear cameras is proposed, and a two-step procedure is presented for estimating its parameters. A camera-centered coordinate frame is defined, and the extrinsic parameters of the camera are determined, that is, the rigid transformation between the calibrating coordinate frame and the camera frame. The accuracy of the calibration method was analyzed through an example application.>

The author gives geometric constructive solutions for 3-D vision problems like positioning a point in space from two views. Using reference points in the scene, no calibration is needed. The method involves only simple geometric computation. From the experiments it is concluded that positioning 3-D points relatively to reference points, is easy and provides more reliable results than absolute positioning as is usually done.>

We prove here that the observation of occluding contours together with a movement of the camera permits the reconstruction of an observed surface. This result is demonstrated under very light constraints, generalizing previous results. Thus, the technique presented can be applied with any movement of the camera, under perspective projection, with surface C1 by parts or even with lines of discontinuity. It allows recovery of the structure of three-dimensional shapes with complex surfaces.

The strengths and the drawbacks of active contour models are described, and the absolute necessity of a criterion for assessing the solutions is pointed out. A method called snake growing, based on successive lengthenings of the snake, is proposed. The strength of this approach is that, at each stage, good convergence conditions are realized and initialization problems can be eliminated.>

International audience

A model-based approach is described for recognizing and locating polyhedral objects from sparse light-stripe data. Based on two scans, consistent interpretations are provided by locating scanned line segments on the faces of a set of known objects. The transformation associated with each interpretation is estimated. Once all the feasible models are transformed to the scene space, a strategy for additional scanning positions is developed to distinguish the object uniquely.>

In this note we present an algorithm for packing spheres in an arbitrary shaped volume. This algorithm is similar to Blum's transform in that it fits spheres into a volume, but it is different in that it fits only tangential spheres, and thereby the data reduction is larger than by Blum's transform. The spheres are of variable radii, which enables us to achieve a hierarchy of intrinsic volume properties, i.e., from gross to more detailed. The result of this algorithm is a graph where the nodes are the centers of spheres and the arcs are the connections between two tangent spheres. Analysis of computational complexity and the time and error considerations are provided.

A refinement is suggested to the O'Rourke-Badler spherical decomposition algorithm which reduces its complexity from 0(n3) to 0(n2)