Abstract The challenge in terrain synthesis for virtual environments is to provide a combination of precise user control over landscape form, with interactive response and visually realistic results. We present a system that builds on parallel pixel‐based texture synthesis to enable interactive creation of an output terrain from a database of heightfield exemplars. We also provide modelers with control over height and surrounding slope by means of constraint points and curves; a paint‐by‐numbers interface for specifying the local character of terrain; coherence controls that allow localization of changes to the synthesized terrain; and copy‐paste functionality to directly transplant terrain regions. Together these contributions provide a level of realism that, based on user experiments, is indistinguishable from real source terrains; user control sufficient for precise placement of a variety of landforms, such as cliffs, ravines and mesas; and synthesis times of 165ms for a 10242 terrain grid.

Motivated by traditional Chinese landscape painting, in which a single brush and a single colour are used to paint a vivid landscape, we propose a new painting interface for volumetric cloud modelling.

Modern laser range scanning campaigns produce extremely large point clouds, and reconstructing a triangulated surface thus requires both out-of-core techniques and significant computational power. We present a GPU-accelerated implementation of the moving least-squares (MLS) surface reconstruction technique. We believe this to be the first GPU-accelerated, out-of-core implementation of surface reconstruction that is suitable for laser range-scanned data. While several previous out-of-core approaches use a sweep-plane approach, we subdivide the space into cubic regions that are processed independently. This independence allows the algorithm to be parallelized using multiple GPUs, either in a single machine or a cluster. It also allows data sets with billions of point samples to be processed on a standard desktop PC. We show that our implementation is an order of magnitude faster than a CPU-based implementation when using a single GPU, and scales well to 8 GPUs.

GPU-based ray casting, as introduced by Krüger and Westermann [2003], is an effective method for volumetric rendering. Unfortunately, conventional methods of Empty Space Skipping (ESS) using spatial partitioning, which accelerate ray casting by culling ray-surface intersection tests in empty parts of the volume, do not align well with GPU architectures. CPUs are usually required for tree generation and parsing, as well as the data transfer from CPU to GPU. Such CPU-based pre-processing is time-consuming, with the result that spatial tree structures are invariably applied to static datasets.

Abstract Curvilinear features extracted from a 2D user‐sketched feature map have been used successfully to constraint a patch‐based texture synthesis of real landscapes. This map‐based user interface does not give fine control over the height profile of the generated terrain. We propose a new texture‐based terrain synthesis framework controllable by a terrain sketching interface. We enhance the realism of the generated landscapes by using a novel patch merging method that reduces boundary artefacts caused by overlapping terrain patches. A more constrained synthesis process is used to produce landscapes that better match user requirements. The high computational cost of texture synthesis is reduced with a parallel implementation on graphics hardware. Our GPU‐accelerated solution provides a significant speedup depending on the size of the example terrain. We show experimentally that our framework is more successful in generating realistic landscapes than current example‐based terrain synthesis methods. We conclude that texture‐based terrain synthesis combined with sketching provides an excellent solution to the user control and realism challenges of virtual landscape generation.

Abstract Traditionally, levels of detail (LOD) for animated characters are computed from a single pose. Later techniques refined this approach by considering a set of sample poses and evaluating a more representative error metric. A recent approach to the character animation problem, animation space, (AS) provides a framework for measuring error analytically. The work presented here uses the animation‐space framework to derive two new techniques to improve the quality of LOD approximations. First, we use an animation‐space distance metric within a progressive mesh‐based LOD scheme, giving results that are reasonable across a range of poses, without requiring that the pose space be sampled. Second, we simplify individual vertices by reducing the number of bones that influence them, using a constrained least‐squares optimization. This influence simplification is combined with the progressive mesh to form a single stream of simplifications. Influence simplification reduces the geometric error by up to an order of magnitude, and allows models to be simplified further than is possible with only a progressive mesh. Quantitative (geometric error metrics) and qualititative (user perceptual) experiments confirm that these new extensions provide significant improvements in quality over traditional, naïve simplification; and while there is naturally some impact on the speed of the off‐line simplification process, it is not prohibitive.

Procedural methods for terrain synthesis are capable of creating realistic depictions of heightfield terrains with little user intervention. However, users often do wish to intervene in controlling the placement and shape of landforms, but without sacrificing realism. In this paper, we present a sketching interface to procedural terrain generation. This system enables users to draw the silhouette, spine and bounding curves of both extruding (hills and mountains) and embedding landforms (river courses and canyons).

Abstract We present a new, single‐rate method for compressing the connectivity information of a connected 2‐manifold triangle mesh with or without boundary. Traditional compression schemes interleave geometry and connectivity coding, and are thus typically unable to utilize information from vertices (mesh regions) they have not yet processed. With the advent of competitive point cloud compression schemes, it has become feasible to develop separate connectivity encoding schemes that can exploit complete, global vertex position information to improve performance. Our scheme demonstrates the utility of this separation of vertex and connectivity coding. By traversing the mesh edges in a consistent fashion, and using global vertex information, we can predict the position of the vertex that completes the unprocessed triangle attached to a given edge. We then rank the vertices in the neighborhood of this predicted position by their Euclidean distance. The distance rank of the correct closing vertex is stored. Typically, these rank values are small, and the set of rank values thus possesses low entropy and compresses very well. The sequence of rank values is all that is required to represent the mesh connectivity—no special split or merge codes are necessary. Results indicate improvements over traditional valence‐based schemes for more regular triangulations. Highly irregular triangulations or those containing a large number of slivers are not well modelled by our current set of predictors and may yield poorer connectivity compression rates than those provided by the best valence‐based schemes.

Abstract We present a simple technique for single‐rate compression of point clouds sampled from a surface, based on a spanning tree of the points. Unlike previous methods, we predict future vertices using both a linear predictor, which uses the previous edge as a predictor for the current edge, and lateral predictors that rotate the previous edge 90°left or right about an estimated normal. By careful construction of the spanning tree and choice of prediction rules, our method improves upon existing compression rates when applied to regularly sampled point sets, such as those produced by laser range scanning or uniform tesselation of higher‐order surfaces. For less regular sets of points, the compression rate is still generally within 1.5 bits per point of other compression algorithms.

Skeletal subspace deformation (SSD), a simple method of character animation used in many applications, has several shortcomings; the best-known being that joints tend to collapse when bent. We present animation space , a generalization of SSD that greatly reduces these effects and effectively eliminates them for joints that do not have an unusually large range of motion.While other, more expensive generalizations exist, ours is unique in expressing the animation process as a simple linear transformation of the input coordinates. We show that linearity can be used to derive a measure of average distance (across the space of poses), and apply this to improving parametrizations.Linearity also makes it possible to fit a model to a set of examples using least-squares methods. The extra generality in animation space allows for a good fit to realistic data, and overfitting can be controlled to allow fitted models to generalize to new poses. Despite the extra vertex attributes, it is possible to render these animation-space models in hardware with no loss of performance relative to SSD.

The task of computer-based free-form shape design is fraught with practical and conceptual difficulties. Incorporating elements of traditional clay sculpting has long been recognized as a means of shielding the user from these complexities. We present warp sculpting, a variant of spatial deformation, which allows deformations to be initiated by the rigid body transformation or uniform scaling of volumetric tools. This is reminiscent of a tool imprinting, flexing, and molding clay. Unlike previous approaches, the deformation is truly interactive. Tools, encoded in a distance field, can have arbitrarily complex shapes. Although individual tools have a static shape, several tools can be applied simultaneously. We enhance the basic formulation of warp sculpting in two ways. First, deformation is toggled to automatically overcome the problem of "sticky" tools, where the object's surface clings to parts of a tool that are moving away. Second, unlike many other spatial deformations, we ensure that warp sculpting remains foldover-free and, hence, prevent self-intersecting objects.

A wavelet-based image-encoding is described which, when used in conjunction with the Difference Engine (a custom-designed VLSI display processor) allows us to reconstruct an image in real-time without the need to set each pixel explicitly. The image is compressed using a quadratic spline-wavelet transform; when reconstructing, an image-adaptive instruction generator attempts to produce the minimal instruction stream to give a good reproduction. The wavelet coefficients are used to decide which regions of the detail images should be retained in the multi-resolution analysis (MRA). A decision is made for each scan-line as to whether it is more economical, in terms of rendering time, to use the 'truncated MRA' or to set the pixels directly. The above approach provides a significant gain over standard image reconstruction/rendering schemes.>