Alex Laier Bordignon

Learning good views through intelligent galleries

The definition of a good view of a 3D scene is highly subjective and strongly depends on both the scene content and the 3D application. Usually, camera placement is performed directly by the user, and that task may be laborious. Existing automatic virtual cameras guide the user by optimizing a single rule, e.g. maximizing the visible silhouette or the projected area. However, the use of a static pre‐defined rule may fail in respecting the users subjective understanding of the scene. This work introduces intelligent design galleries, a learning approach for subjective problems such as the camera placement. The interaction of the user with a design gallery teaches a statistical learning machine. The trained machine can then imitate the user, either by pre‐selecting good views or by automatically placing the camera. The learning process relies on a Support Vector Machines for classifying views from a collection of descriptors, ranging from 2D image quality to 3D features visibility. Experiments of the automatic camera placement demonstrate that the proposed technique is efficient and handles scenes with occlusion and high depth complexities. This work also includes user validations of the intelligent gallery interface.

Statistical optimization of octree searches

This work emerged from the following observation: usual search procedures for octrees start from the root to retrieve the data stored at the leaves. But as the leaves are the farthest nodes to the root, why start from the root? With usual octree representations, there is no other way to access a leaf. However, hashed octrees allow direct access to any node, given its position in space and its depth in the octree. Search procedures take the position as an input, but the depth remains unknown. This work proposes to estimate the depth of an arbitrary node through a statistical optimization of the average cost of search procedures. As the highest costs of these algorithms are obtained when starting from the root, this method improves on both the memory footprint by the use of hashed octrees, and execution time through the proposed optimization.

Exploratory visualization based on multidimensional transfer functions and star coordinates

Exploration and analysis of multivariate data play an important role in different domains. This work proposes a simple interface prototype that allows a human user to visually explore multivariate spatial objects, such as images, sequence of images or volume. It uses star coordinates as a widget to display the multivariate data on the computer 2D screen. The user then identifies a feature on this powerful coordinate system by mapping a selected feature region on that widget to a color and opacity. As a visual result, the feature is rendered on the objects space composing the use of this maps and the star coordinates projection. Some examples illustrate the potential of this interface

Point set compression through BSP quantization

This work introduces a new compression scheme for point sets. This scheme relies on an adaptive binary space partition (BSP) which takes into account the geometric structure of the point set. This choice introduces geometrical rather than combinatorial information in the compression scheme. In order to effectively improve the final compression ratio, this partition is encoded in a progressive manner, decreasing the number of bits used for the quantisation at each subdivision. This strategy distributes the extra cost of the geometry encoding onto the maximal number of points, compressing in average 15% more than previous techniques

Tuning Manifold Harmonics Filters

There are several techniques for automatic music visualization, which are included with virtually any media player. The basic ingredient of those techniques is spectral analysis of the sound, used to automatically generate parameters for procedural image generation. However, only a few music visualizations rely on 3d models. This paper proposes to use spectral mesh processing techniques, namely manifold harmonics, to produce 3d music visualization. The images are generated from 3d models by deforming an initial shape, mapping the sound frequencies to the mesh harmonics. A concise representation of such frequency mapping is proposed to permit for an animated gallery interface with genetic reproduction. Such galleries allow the user to quickly navigate between visual effects. Rendering such animated galleries in real-time is a challenging task, since it requires computing and rendering the deformed shapes at a very high rate. This paper introduces a direct GPU implementation of manifold harmonics filters, which allows to display animated gallery.

Technical Section: Point-based rendering of implicit surfaces in R4

We present a point-based algorithm for rendering implicit surfaces in R4. Our algorithm combines a new method for approximating an implicit surface with points that uses interval arithmetic for topological robustness with a new 4D illumination model that together with a color transfer function enhance the visualization of a 2-dimensional surface in 4-dimensional space. We also discuss a GPU implementation of our algorithm.

Geometry Super-Resolution by Example

The acquisition of high-resolution 3D models still requires delicate and time-consuming processes. In particular, each detail of the object should be scanned separately, although they may be similar. This can be simplified by copying a small set of details at different places of the model, synthesizing high geometric resolution from details exemplars, as introduced in this paper for three different contexts : when the detail exemplars are scanned separately at high resolution, when they are synthesized or edited from other models, or when they are obtained by accumulating repeated instances of the detail in the low-resolution scan. The main challenge here is to correctly register the high-resolution details with the low resolution model. To address this issue, this work proposes a careful resolution manipulation of 3D scans at each step of an automatic registration pipeline, combined with a robust selection of alignments. This results in a fully automatic process for geometry super-resolution by example. Experiments on synthetic and real data sets show applicability in different contexts, including resolution increase, noise removal by example and geometric texture insertion.

Scale-Space for Union of 3D Balls

Shape discretization through union of weighted points or balls appears as a common representation in different fields of computer graphics and geometric modeling. Among others, it has been very successful for implicit surface reconstruction with radial basis functions, molecular atomic models, fluid simulation from particle systems and deformation tracking with particle filters. These representations are commonly generated from real measurements or numerical computations, which may require filtering and smoothing operations.This work proposes a smoothing mechanism for union of balls that tries to inherit from the scale-space properties of bi-dimensional curvature motion: it avoids disconnecting the shape, prevents self-intersection, regularly decreases the area and convexifies the shape. The smoothing is computed iteratively by moving each ball of the union according to a combination of projected planar curvature motions. Experiments exhibits nice properties of this scale-space.

Approximations by Smooth Transitions in Binary Space Partitions

This work proposes a simple approximation scheme for discrete data that leads to an infinitely smooth result without global optimization. It combines the flexibility of binary space partitions trees with the statistical robustness of smooth transition regression trees. The construction of the tree is straightforward and easily controllable, using error-driven metrics or external constraints. Moreover, it leads to a concise representation. Applications on synthetic and real data, both scalar and vector-valued demonstrated the effectiveness of this approach.

Support Vectors Learning for Vector Field Reconstruction

Sampled vector fields generally appear as measurements of real phenomena. They can be obtained by the use of a Particle Image Velocimetry acquisition device, or as the result of a physical simulation, such as a fluid flow simulation, among many examples. This paper proposes to formulate the unstructured vector field reconstruction and approximation through Machine-Learning. The machine learns from the samples a global vector field estimation function that could be evaluated at arbitrary points from the whole domain. Using an adaptation of the Support Vector Regression method for multi-scale analysis, the proposed method provides a global, analytical expression for the reconstructed vector field through an efficient non-linear optimization. Experiments on artificial and real data show a statistically robust behavior of the proposed technique.