Jianning Wang

A hole-filling strategy for reconstruction of smooth surfaces in range images

Creating models of real scenes is a complex task for which the use of traditional modelling techniques is inappropriate. For this task, laser rangefinders are frequently used to sample the scene from several viewpoints, with the resulting range images integrated into a final model. In practice, due to surface reflectance properties, occlusions and accessibility limitations, certain areas of the scenes are usually not sampled, leading to holes and introducing undesirable artifacts in the resulting models. We present an algorithm for filling holes on surfaces reconstructed from point clouds. The algorithm is based on moving least squares and can recover both geometry and shading information, providing a good alternative when the properties to be reconstructed are locally smooth. The reconstruction process is mostly automatic and the sampling rate in the reconstructed areas follows the given samples. We demonstrate the use of the algorithm on both real and synthetic data sets to obtain complete geometry and reasonable shading.

Filling holes on locally smooth surfaces reconstructed from point clouds

Abstract Creating models of real scenes is a complex task for which the use of traditional modeling techniques is inappropriate. For this task, laser rangefinders are frequently used to sample the scene from several viewpoints, with the resulting range images integrated into a final model. In practice, due to surface reflectance properties, occlusions and accessibility limitations, certain areas of the scenes are usually not sampled, leading to holes and introducing undesirable artifacts in the resulting models. We present an algorithm for filling holes on surfaces reconstructed from point clouds. The algorithm is based on moving least squares and can interpolate both geometry and shading information. The reconstruction process is mostly automatic and the sampling rate of the given samples is preserved in the reconstructed areas. We demonstrate the use of the algorithm on both real and synthetic datasets to obtain complete geometry and plausible shading.

Piecewise C/sup 1/ continuous surface reconstruction of noisy point clouds via local implicit quadric regression

This paper addresses the problem of surface reconstruction of highly noisy point clouds. The surfaces to be reconstructed are assumed to be 2-manifolds of piecewise C/sup 1/ continuity, with isolated small irregular regions of high curvature, sophisticated local topology or abrupt burst of noise. At each sample point, a quadric field is locally fitted via a modified moving least squares method. These locally fitted quadric fields are then blended together to produce a pseudo-signed distance field using Shepards method. We introduce a prioritized front growing scheme in the process of local quadrics fitting. Flatter surface areas tend to grow faster. The already fitted regions will subsequently guide the fitting of those irregular regions in their neighborhood.

Improved Scene Reconstruction from Range Images

The modeling of real scenes is a complex and challenging task for which the use of laser rangefinders is one of the most promising approaches. Unfortunately, in many situations, it is not possible or practical to guarantee appropriate sampling of all surfaces in the scene. For example, occlusions and accessibility limitations to certain regions of the scene may cause some areas not to be visible by the scanner, resulting in incomplete or incorrectly reconstructed models. This paper describes a pipeline and a system implementation for improving model reconstruction from incomplete information available from range images.

3D Surface Reconstruction from Endoscopic Videos

Endoscopy is a popular procedure to help surgeons investigate the interior of a patient’s organ and find abnormal regions (e.g., polyps). However, it requires a great expertise using only a stream of 2D images of the interior, and there is a possibility for the physicians to miss some polyps. In this case, a 3D reconstruction of the interior surface of the organ will be very helpful. It turns the stream of 2D images into a meaningful 3D model. The physicians could then spend more time scrutinizing the interior surface. In addition, the reconstruction result will provides more details about the patient’s organ including a coordinate system and could be saved for later uses. Shape from X (X={shading, motion, texture, etc.}) techniques, which have been studied for decades in the computer vision community, are good candidates for this purpose. However, the specific problems associated with reconstruction from endoscopic video make traditional shape from X techniques inappropriate:

Reconstructing manifold and non-manifold surfaces from point clouds

This paper presents a novel approach for surface reconstruction from point clouds. The proposed technique is general in the sense that it naturally handles both manifold and non-manifold surfaces, providing a consistent way for reconstructing closed surfaces as well as surfaces with boundaries. It is also robust in the presence of noise, irregular sampling and surface gaps. Furthermore, it is fast, parallelizable and easy to implement because it is based on simple local operations. In this approach, surface reconstruction consists of three major steps: first, the space containing the point cloud is subdivided, creating a voxel representation. Then, a voxel surface is computed using gap filling and topological thinning operations. Finally, the resulting voxel surface is converted into a polygonal mesh. We demonstrate the effectiveness of our approach by reconstructing polygonal models from range scans of real objects as well as from synthetic data.

Surface reconstruction using oriented charges

We introduce the notion of oriented charges to compute distance fields from unorganized point clouds. Unlike traditional implicit function methods that usually require normal information, our approach only uses information about the position of the samples. Our approach is adaptive, using an octree to reorganize the input points, thus avoiding difficulties often associated with the use of local distance fields. The resulting representation is simple, efficient and general in the sense that it handles complex models of arbitrary topology, is robust in the presence of noise and fills holes automatically.

Reconstructing regular meshes from points: A parameterization-based approach

We propose an algorithm for reconstructing regular meshes from unorganized point clouds. At first, a nearly isometric point parameterization is computed using only the location of the points. A mesh, composed of nearly equilateral triangles, is later created using a regular sampling pattern. This approach produces meshes with high visual quality and suitable for use with applications such as finite element analysis, which tend to impose strong constraints on the regularity of the input mesh. Geometric properties, such as local connectivity and surface features, are identified directly from the points and are stored independent of the resulting mesh. This decoupling preserves most details and allows more flexibility for meshing. The resulting parameterization supports several direct applications, such as texturing and bump mapping. In addition, novel boundary identification and cut parameterization algorithms are proposed to overcome the difficulties caused by cuts, non-closed surfaces and possible self-overlapping parameter patches. We demonstrate the effectiveness of our approach by reconstructing regular meshes from real datasets, such as a human colon obtained from CT scan and objects digitized using laser scanners.

Per-Pixel Camera Calibration for 3D Range Scanning

Structured-light rangefinder is distinguished from other range scanning systems by its use of off-the-shelf hardware and fast data acquisition. We propose a novel approach to calibrate such a system, namely, calibrate the camera and the projector of the system. This approach handles all types of distortions and produces results in high accuracy. Basically, camera calibration techniques compute the pixel-ray correspondences and represent them by a mathematic model with limited number of parameters. These parametric models, however, cannot model general distortions while distortions not modelled, sometimes, can greatly affect the quality of further applications. The proposed approach computes and maintains a ray database for all pixels explicitly. By trading memory for accuracy, this approach solves the above problem. The resulting ray database can be directly used for 3D point reconstruction using calibration system. It is also useful for model fitting and other operations.