Search

This paper presents a novel method for multi-view sequential scene reconstruction scenarios such as in aerial video, that exploits the constraints imposed by the path of a moving camera to allow for a new way of detecting and correcting inaccuracies in the feature tracking and structure computation processes. The main contribution of this paper is to show that for short, planar segments of a continuous camera trajectory, parallax movement corresponding to a viewed scene point should ideally form a scaled and translated version of this trajectory when projected onto a parallel plane. This creates two constraints, which differ from those of standard factorization, that allow for the detection and correction of inaccurate feature tracks and to improve scene structure. Results are shown for real and synthetic aerial video and turntable sequences, where the proposed method was shown to correct outlier tracks, detect and correct tracking drift, and allow for a novel improvement of scene structure, additionally resulting in an improved convergence for bundle adjustment optimization.

An algorithm that shows how ray divergence in multi-view stereo scene reconstruction can be used towards improving bundle adjustment weighting and conditioning is presented. Starting with a set of feature tracks, ray divergence when attempting to compute scene structure for each track is first obtained. Assuming accurate feature matching, ray divergence reveals mainly camera parameter estimation inaccuracies. Due to its smooth variation across neighboring feature tracks, from its histogram a set of weights can be computed that can be used in bundle adjustment to improve its convergence properties. It is proven that this novel weighting scheme results in lower reprojection errors and faster processing times than others such as image feature covariances, making it very suitable in general for applications involving multi-view pose and structure estimation.

This paper introduces a non-parametric sequential frame decimation algorithm for image sequences in low-memory streaming environments. Frame decimation reduces the number of input frames to increase pose and structure robustness in Structure and Motion (SaM) applications. The main contribution of this paper is the introduction of a sequential low-memory work-flow for frame decimation in embedded systems where memory and memory traffic come at a premium. This approach acts as an online preprocessing filter by removing frames that are ill-posed for reconstruction before streaming. The introduced sequential approach reduces the number of needed frames in memory to three in contrast to global frame decimation approaches that use at least ten frames in memory and is therefore suitable for low-memory streaming environments. This is moreover important in emerging systems with large format cameras which acquire data over several hours and therefore render global approaches impossible. In this paper a new decimation metric is designed which facilitates sequential keyframe extraction fit for reconstruction purposes, based on factors such as a correspondence-to-feature ratio and residual error relationships between epipolar geometry and homography estimation. The specific design of the error metric allows a local sequential decimation metric evaluation and can therefore be used on the fly. The approach has been tested with various types of input sequences and results in reliable low-memory frame decimation robust to different frame sampling frequencies and independent of any thresholds, scene assumptions or global frame analysis.

A correspondence and camera error analysis for dense correspondence applications such as structure from motion is introduced. This provides error introspection, opening up the possibility of adaptively and progressively applying more expensive correspondence and camera parameter estimation methods to reduce these errors. The presented algorithm evaluates the given correspondences and camera parameters based on an error generated through simple triangulation. This triangulation is based on the given dense, non-epipolar constraint, correspondences and estimated camera parameters. This provides an error map without requiring any information about the perfect solution or making assumptions about the scene. The resulting error is a combination of correspondence and camera parameter errors. An simple, fast low/high pass filter error factorization is introduced, allowing for the separation of correspondence error and camera error. Further analysis of the resulting error maps is applied to allow efficient iterative improvement of correspondences and cameras.

We present two algorithms for data compression and occlusion culling that im- prove interactive, adaptive isosurface extraction from large volume datasets. Our algorithm, based on hierarchical tetrahedral meshes defined by longest edge bisection, allows arbitrary isosurfaces to be adaptively extracted at interactive rates from losslessly compressed volumes where the region of interest, determined at runtime by user interaction, is decompressed on- the-fly. For interactive applications, we exploit frame-to-frame coherence between consecutive views to simplify the mesh structure in occluded regions and eliminate occluded triangles sig- nificantly reducing the complexity of the visualized surface and the underlying multiresolution volume representation. We extend the use of hardware accelerated occlusion queries to adap- tive isosurface extraction applications where the surface geometry and topology change with the level-of-detail and view-point and the user can select an arbitrary isovalue for visualization.

Multiresolution methods provide a means for representing data at multiple levels of detail. They are typically based on a hierarchical data organization scheme and update rules needed for data value computation. We use a data organization that is based on what we call \(\root n \of 2 \) subdivision, where n is the dimension of the data set. The main advantage of \(\root n \of 2 \) subdivision, compared to quadtree (n = 2) or octree (n = 3) organizations, is that the number of vertices is only doubled in each subdivision step instead of multiplied by a factor of 2 n , i. e., four or eight, respectively. To update data values we use n-variate B-spline wavelets, which yield better approximations for each level of detail. We develop a lifting scheme for n = 2 and n = 3 based on the \(\root n \of 2 \)-subdivision scheme. We obtain narrow masks that provide a basis for out-of-core techniques as well as view-dependent visualization and adaptive, localized refinement.

We introduce a novel wavelet decomposition algorithm that makes a number of powerful new surface design operations practical. Wavelets, and hierarchical representations generally, have held promise to facilitate a variety of design tasks in a unified way by approximating results very precisely, thus avoiding a proliferation of undergirding mathematical representations. However, traditional wavelet decomposition is defined from fine to coarse resolution, thus limiting its efficiency for highly precise surface manipulation when attempting to create new non-local editing methods.

Currently, large physics simulations produce 3D discretized field data whose individual isosurfaces, after conventional extraction processes, contain upwards of hundreds of millions of triangles. Detailed interactive viewing of these surfaces requires (a) powerful compression to minimize storage, and (b) fast view-dependent optimization of display triangulations to most effectively utilize high-performance graphics hardware. In this work, we introduce the first end-to-end multiresolution dataflow strategy that can effectively combine the top performing subdivision-surface wavelet compression and view-dependent optimization methods, thus increasing efficiency by several orders of magnitude over conventional processing pipelines. In addition to the general development and analysis of the dataflow, we present new algorithms at two steps in the pipeline that provide the “glue” that makes an integrated large-scale data visualization approach possible. A shrink-wrapping step converts highly detailed unstructured surfaces of arbitrary topology to the semi-structured meshes needed for wavelet compression. Remapping to triangle bintrees minimizes disturbing “pops” during realtime displaytriangulation optimization and provides effective selective-transmission compression for out-of-core and remote access to extremely large surfaces. Overall, this is the first effort to exploit semi-structured surface representations for a complete large-data visualization pipeline.

We present a new method for constructing multiresolution representations of data sets that contain material interfaces. Material interfaces embedded in the meshes of computational data sets are often a source of error for simplification algorithms because they represent discontinuities in the scalar or vector field over mesh elements. By representing material interfaces explicitly, we are able to provide separate field representations for each material over a single cell. Multiresolution representations utilizing separate field representations can accurately approximate datasets that contain discontinuities without placing a large percentage of cells around the discontinuous regions. Our algorithm uses a multiresolution tetrahedral mesh supporting fast coarsening and refinement capabilities; error bounds for feature preservation; explicit representation of discontinuities within cells; and separate field representations for each material within a cell.