Ruwen Schnabel

Efficient RANSAC for Point-Cloud Shape Detection

In this paper we present an automatic algorithm to detect basic shapes in unorganized point clouds. The algorithm decomposes the point cloud into a concise, hybrid structure of inherent shapes and a set of remaining points. Each detected shape serves as a proxy for a set of corresponding points. Our method is based on random sampling and detects planes, spheres, cylinders, cones and tori. For models with surfaces composed of these basic shapes only, for example, CAD models, we automatically obtain a representation solely consisting of shape proxies. We demonstrate that the algorithm is robust even in the presence of many outliers and a high degree of noise. The proposed method scales well with respect to the size of the input point cloud and the number and size of the shapes within the data. Even point sets with several millions of samples are robustly decomposed within less than a minute. Moreover, the algorithm is conceptually simple and easy to implement. Application areas include measurement of physical parameters, scan registration, surface compression, hybrid rendering, shape classification, meshing, simplification, approximation and reverse engineering.

Octree-based point-cloud compression

In this paper we present a progressive compression method for point sampled models that is specifically apt at dealing with densely sampled surface geometry. The compression is lossless and therefore is also suitable for storing the unfiltered, raw scan data. Our method is based on an octree decomposition of space. The point-cloud is encoded in terms of occupied octree-cells. To compress the octree we employ novel prediction techniques that were specifically designed for point sampled geometry and are based on local surface approximations to achieve high compression rates that outperform previous progressive coders for point-sampled geometry. Moreover we demonstrate that additional point attributes, such as color, which are of great importance for point-sampled geometry, can be well integrated and efficiently encoded in this framework.

Technical Section: Robust normal estimation for point clouds with sharp features

This paper presents a novel technique for estimating normals on unorganized point clouds. Methods from robust statistics are used to detect the best local tangent plane for each point. Therefore the algorithm is capable to deal with points located in high curvature regions or near/on complex sharp features, while being highly robust with respect to noise and outliers. In particular, the presented method reliably recovers sharp features but does not require tedious manual parameter tuning as done by current methods. The key ingredients of our approach are a robust noise-scale estimator and a kernel density estimation (KDE) based objective function. In contrast to previous approaches the noise-scale estimation is not affected by sharp features and achieves high accuracy even in the presence of outliers. In addition, our normal estimation procedure allows detection and elimination of outliers. We confirm the validity and reliability of our approach on synthetic and measured data and demonstrate applications to point cloud denoising.

Detail-preserving surface inpainting

Inpainting is a well-known technique in the context of image and art restoration, where paint losses are filled up to the level of the surrounding paint and then coloured to match. Analogue tasks can be found in 3D geometry processing, as digital representations of real-world objects often contain holes, due to hindrances during data acquisition or as a consequence of interactive modelling operations. In this paper we present a novel approach to automatically fill-in holes in structured surfaces where smooth hole filling is not sufficient. Previous approaches inspired by texture synthesis algorithms require specific spatial structures to identify holes and possible candidate fragments to be copied to defective regions. Consequently, the results depend heavily on the choice and location of these auxiliary structures, such that for instance symmetries are not reconstructed faithfully. In contrast, our approach is based on local neighbourhoods and therefore insensitive with respect to similarity transformations. We use so-called guidance surfaces to guide and prioritise the atomic filling operations, such that even non-trivial and larger holes can be filled consistently. The guidance surfaces are automatically computed and iteratively updated during the filling process, but can also incorporate any additional information about the surface, if available.

Patch-based texture interpolation

In this paper, we present a novel exemplar‐based technique for the interpolation between two textures that combines patch‐based and statistical approaches. Motivated by the notion of texture as a largely local phenomenon, we warp and blend small image neighborhoods prior to patch‐based texture synthesis. In addition, interpolating and enforcing characteristic image statistics faithfully handles high frequency detail. We are able to create both intermediate textures as well as continuous transitions. In contrast to previous techniques computing a global morphing transformation on the entire input exemplar images, our localized and patch‐based approach allows us to successfully interpolate between textures with considerable differences in feature topology for which no smooth global warping field exists.

Towards semantic scene analysis with time-of-flight cameras

For planning grasps and other object manipulation actions in complex environments, 3D semantic information becomes crucial. This paper focuses on the application of recent 3D Time-of-Flight (ToF) cameras in the context of semantic scene analysis. For being able to acquire semantic information from ToF camera data, we a) pre-process the data including outlier removal, filtering and phase unwrapping for correcting erroneous distance measurements, and b) apply a randomized algorithm for detecting shapes such as planes, spheres, and cylinders. We present experimental results that show that the robustness against noise and outliers of the underlying RANSAC paradigm allows for segmenting and classifying objects in 3D ToF camera data captured in natural mobile manipulation setups.

Effective Visualization of Short Routes

In this work we develop a new alternative to conventional maps for visualization of relatively short paths as they are frequently encountered in hotels, resorts or museums. Our approach is based on a warped rendering of a 3D model of the environment such that the visualized path appears to be straight even though it may contain several junctions. This has the advantage that the beholder of the image gains a realistic impression of the surroundings along the way which makes it easy to retrace the route in practice. We give an intuitive method for generation of such images and present results from user studies undertaken to evaluate the benefit of the warped images for orientation in unknown environments.

Special Section: Point-Based Graphics: Fast vector quantization for efficient rendering of compressed point-clouds

We present a point-cloud compression algorithm that allows fast parallel decompression on the GPU suitable for interactive applications. The algorithm is based on vector quantization of an atlas of height-fields that have been sampled over primitive shapes which approximate the geometry. We introduce novel vector quantization acceleration techniques to facilitate fast compression as well. We achieve bitrates of less than four bits per normal-equipped point. Our method enables hole-free level-of-detail point rendering. We also show that using only up to two bits per point, high-quality renderings can still be obtained if normals are estimated in image-space. Even lower bitrates are obtained for storage on disk if arithmetic coding is used.

A Parallelly Decodeable Compression Scheme for Efficient Point-Cloud Rendering

We present a point-cloud compression algorithm that allows fast parallel decompression on the GPU for interactive applications. We achieve bitrates of less than four bits per normal-equipped point. Our method enables holefree level-of-detail point rendering. We also show that using only up to two bits per point, high-quality renderings can still be obtained if normals are estimated in image-space. The algorithm is based on vector quantization of an atlas of height-fields that have been sampled over primitive shapes which approximate the geometry.

Variational Surface Approximation and Model Selection

We consider the problem of approximating an arbitrary generic surface with a given set of simple surface primitives. In contrast to previous approaches based on variational surface approximation, which are primarily concerned with finding an optimal partitioning of the input geometry, we propose to integrate a model selection step into the algorithm in order to also optimize the type of primitive for each proxy. Our method is a joint global optimization of both the partitioning of the input surface as well as the types and number of used shape proxies. Thus, our method performs an automatic trade‐off between representation complexity and approximation error without relying on a user supplied predetermined number of shape proxies. This way concise surface representations are found that better exploit the full approximative power of the employed primitive types.