Academic
Publications
Pose Priors for Simultaneously Solving Alignment and Correspondence

Pose Priors for Simultaneously Solving Alignment and Correspondence,10.1007/978-3-540-88688-4_30,Francesc Moreno-noguer,Vincent Lepetit,Pascal Fua

Pose Priors for Simultaneously Solving Alignment and Correspondence   (Citations: 10)
BibTex | RIS | RefWorks Download
Estimating a camera pose given a set of 3D-object and 2D-image feature points is a well understood problem when correspondences are given. However, when such correspondences cannot be established a priori, one must simultaneously compute them along with the pose. Most current approaches to solving this problem are too computationally intensive to be practical. An interesting exception is the SoftPosit algorithm, that looks for the solution as the minimum of a suitable objective function. It is arguably one of the best algorithms but its iterative nature means it can fail in the presence of clutter, occlusions, or repetitive patterns. In this paper, we propose an approach that overcomes this limitation by taking advantage of the fact that, in practice, some prior on the camera pose is often available. We model it as a Gaussian Mixture Model that we progressively refine by hypothesizing new correspondences. This rapidly reduces the number of potential matches for each 3D point and lets us explore the pose space more thoroughly than SoftPosit at a similar computational cost. We will demonstrate the superior performance of our approach on both synthetic and real data.
Conference: European Conference on Computer Vision - ECCV , pp. 405-418, 2008
Cumulative Annual
View Publication
The following links allow you to view full publications. These links are maintained by other sources not affiliated with Microsoft Academic Search.
    • ...This joint scheme is well studied for the camera pose estimation using rigid objects [8, 23] but it has not been studied for the deformable surface reconstruction problem...

    Appu Shajiet al. Simultaneous point matching and 3D deformable surface reconstruction

    • ...Inspired on a recent paper on rigid object pose estimation [18], we reduce the complexity of this problem using weak priors on pose and shape, that we learn from training data, and model as Gaussian Mixture Models...
    • ...Indeed, several works have already been proposed for retrieving matches and the six degrees of freedom of the pose in rigid objects [6, 10, 18, 20]...
    • ...Many variants of this strategy have been proposed over the years to reduce the computational cost [6, 10, 18, 20]...
    • ...In fact, our approach is inspired in [18], a recent work that uses priors on the camera pose to register rigid objects...
    • ...Pose priors are computed following a similar methodology as in [18]...
    • ...In the synthetic results we compare our approach to [18], which we denote by BPnP...
    • ...In order to evaluate our approach we proceeded similarly as in [18]...
    • ...Figure 5. Comparing the convergence rate of our approach and BPnP [18]...

    Jordi Sanchez-Rieraet al. Simultaneous pose, correspondence and non-rigid shape

    • ...We build upon the so-called Blind PnP approach [4] that was designed to simultaneously establish 2D to 3D correspondences and estimate camera pose...
    • ...This is addressed in the Blind PnP [4] by introducing weak pose priors, that constrain where the camera can look at, and guide the search for correspondences...

    Eduard Serradellet al. Combining Geometric and Appearance Priors for Robust Homography Estima...

    • ...However, in addition to being restricted to localizing only 3D object with four or more non-coplanar points, certain configuration of data or situations with large amount of clutter and occlusions might still cause the SoftPOSIT algorithm to fail [24]...

    Vicky Wonget al. A combined particle filter and deterministic approach for underwater o...

    • ...Incorporating probabilistic predictions into RANSAC has been previously investigated in [15] for the case of weak priors and also in an EKF context [28]...

    Javier Civeraet al. 1-point RANSAC for EKF-based Structure from Motion

Sort by: