Martin Madaras

Skeleton texture mapping

In the article an idea for a novel way of mapping of textures onto a surface of 3D model is introduced. Our technique is based on two interlocking mappings. The first one maps surface vertices onto a computed skeleton and the second one maps the surrounding area of each skeleton segment into a rectangle with size based on the surface properties around the segment. Furthermore, these rectangles are packed into a squared texture called skeleton texture map (STM) by approximately solving a palette loading problem. Our technique enables the mapping of a texture onto the surface without necessity to store texture coordinates with the model data and it is also suitable for surfaces with a topology non-homotopic to a sphere with higher order genus and unlimited structure branching.

Physically inspired stretching for skinning animation of non-rigid bodies

We propose a physically inspired stretching model for non-rigid bodies with linear skeletons. Given an input model composed of linear skeleton segments, it extract scaling matrices that can be directly used in skinning animation. The stretching model evaluates stretching of the body cause by gravitational force and stretching of the body caused by muscle contraction. Our model is based on small deformation theory, which can be directly applied on cylindrical shapes. Since the input body may differ from a cylindrical shape, it is decomposed into several cylindrical parts and stretching factors are calculated for each part individually. Next, the body is stretched along the skeleton based on the function derived from the sum of skeleton curvature. Finally, a system for the visualization of a particle-based simulation using linear blend skinning is created and enhanced with out stretching model.

Controllable Skeleton-Sheets Representation Via Shape Diameter Function

The Shape Diameter Function (SDF) is a scalar function defined on the mesh surface expressing a measure of the diameter of the objects volume in the neighborhood of each point on the surface. In our approach we propose to compute internal skeleton, from which every mesh vertex is visible. There are automatic techniques for curve-skeleton extraction, but extracted curve-skeletons does not satisfy this condition. Therefore, we have extended Laplacian smoothing based method via the SDF term that moves skeleton vertices near medial axis. In addition, SDF values can be used to modify parameters as weights and grouping distance during Laplacian-based skeleton extraction to obtain better results with skeleton extraction. Resulting skeleton is reliable, centered in the mesh volume, consist of curves and sheets and is useful for skeleton based parametrization of meshes. Our approach is designed to work on generalized discrete geometry data, particularly point clouds, by estimation of Laplacian from local triangulations.

Skeleton-based 3D Surface Parameterization Applied on Texture Mapping

Abstract Assume a 2D manifold surface topologically equivalent to a sphere with handles we propose a novel 3D surface parametrization along the surface skeleton. First, we use a global mapping of the surface vertices onto a computed skeleton. Second, we use local mapping of the surrounding area of each skeleton segment into a small rectangle whose size is derived based on the surface properties around the segment. Each rectangle can be textured by assigning the local u;v texture coordinates. Furthermore, these rectangles are packed into a large squared texture called skeleton texture map (STM) by approximately solving a palette loading problem. Our technique enables the mapping of a texture onto the surface without necessity to store texture coordinates together with the model data. In other words it is enough to store the geometry data with STM and the coordinates are calculated on the fly.

Optical-inertial Synchronization of MoCap Suit with Single Camera Setup for Reliable Position Tracking.

We propose a method for synchronization of an inertial motion capture suit and a single camera optical setup. Proposed synchronization is based on an iterative optimization of an energy potential in image space, minimizing the error between the camera image and a rendered virtual representation of the scene. For each frame, an input skeleton pose from the mocap suit is used to render a silhouette of a subject. Moreover, the local neighborhood around the last known position is searched by matching the silhouette to the distance transform representation of the camera image based on Chamfer matching. Using the combination of the camera tracking and the inertial motion capture suit, it is possible to retrieve the position of the joints that are hidden from the camera view. Moreover, it is possible to capture the position even if it cannot be captured by the suit sensors. Our system can be used for both real-time tracking and off-line post-processing of already captured mocap

Skeletex: Skeleton-texture Co-representation for Topology-driven Real-time Interchange and Manipulation of Surface Regions

Mesh processing algorithms depend on quick access to the local neighborhood, which requires costly memory queries. Moreover, even having access to the local neighborhood is not enough to efficiently perform many geometry processing algorithms in an automatic or semi‐automatic way. As humans, we often imagine mesh editing at the level of topological information, e.g., altering surface features, adding limbs, etc., which is not supported by current data structures. These limitations come from the widely used mesh representations because the needed information is not implicitly defined by the structure. We propose a novel model representation called Skeletex. Each 3D model is decomposed into two elements: a skeletal structure that encodes the model topology and a vector displacement map to capture fine details of the geometry. Such a co‐representation contains the topology information, as well as the information about the local vertex neighborhood at each texel. Additionally, our data structure facilitates an automatic skeleton‐based cross‐parameterization. This allows us to implement the mesh manipulation tasks in parallel, using a unified streamlined pipeline that directly maps to the GPU. We demonstrate the capabilities of our data structure by implementing surface region transfer and mesh morphing of 3D models.

Fast parallel computation of shape diameter function

We outline and compare two approaches for the computation of the shape diameter function (SDF) on the GPU. The SDF is a scalar function describing the local thickness of an object. It can be used for consistent mesh partitioning and skeletonization. In the first approach, we have reorganized the tracing of rays to be well suited for the rasterization hardware. To the best of our knowledge, this is the first method to show how to compute the SDF using only the rasterization hardware and without the need of any acceleration data structures. The second approach uses parallel ray casting and octree traversal using OpenCL. We demonstrate that first method achieves similar results as the ray casting using OpenCL, it is faster for large meshes and it simpler to implement. Furthermore, we extend the SDF computation by fast post-processing using texture-space diffusion. Fast SDF computation can be use in many applications such as automatic skeleton extraction demonstrated in the article.

Skeleton-based matching for animation transfer and joint detection

In this paper we present a new algorithm for establishing correspondence between objects based on matching of extracted skeletons. First, a point cloud of an input model is scanned. Second, a skeleton is extracted from the scanned point cloud. In the last step, all the extracted skeletons are matched based on valence of vertices and segment lengths. The matching process yields into two direct applications - topological mapping and segment mapping. Topological mapping can be used for detection of joint positions from multiple scans of articulated figures in different poses. Segment mapping can be used for animation transfer and for transferring of arbitrary surface per-vertex properties. Our approach is unique, because it is based on matching of extracted skeletons only and does not require vertex correspondence.

Skeleton-based Joints Position Detection

We present a system for detection of joint positions in scans of articulated models. Our method is based purely on skeletons extracted from scanned point clouds of input models. First, skeletons are extracted from scans and then an estimation of possible matches between skeletons is performed. The matches are evaluated and sorted out. The whole matching process is fully automatic, but some user-driven suggestions can be included. Finally, we pick the best matching of skeletons and create a union-skeleton containing all the nodes from all the skeletons. We find nodes in the union-skeleton with rotation changes higher than the predefined threshold. We take these nodes as joints and visualize them in original scans.

Parallelization of Mesh Contraction and Fairing using OpenCL

We propose a parallel method for computing local Laplacian curvature flows for triangular meshes. Laplace operator is widely used in mesh processing for mesh fairing, noise removal or curvature estimation. If the Laplacian flow is used in global sense constraining a whole mesh with an iterative weighted linear system, it can be used even for mesh contraction. However, numerical solution of such a global linear system is computationally expensive. Therefore, we have developed a method to compute such an iterative linear system using only local neighbourhoods of each vertex in parallel. Parallel computation of local linear systems is performed on GPU using OpenCL. We have evaluated speedups of the parallelization using both local and global Laplacian flows. We show test cases, where the parallel local method can be used for mesh fairing. In contrary, we also investigate and outline a fail case, where the local Laplacian flow cannot be used. When the local Laplacian flow has problems with global convergence, we offer a global parallelization of the linear system solving as an alternative.