Israel SIGGRAPH meeting on March 26

Israel SIGGRAPH meeting on March 26, 1999

Chair: Anath Fischer
            Laboratory for Computer Graphics and CAD
            Dept. of Mechanical Engineering
            Technion - Israel Institute of Technology

The following program is also available in PostScript format. The PostScript file is to be printed double-sided on A4 paper, and folded into three columns with INVITATION and the digit "6" on the exposed columns.

Time Speaker Title Abstract

8:30 REFRESHMENTS

9:00 Marc Alexa
Computer Science Dept.
Darmstadt
Univ. of Technology, Germany Morphing Polyhedral Shapes with Scattered Features Metamorphosis techniques for images and volume models have gained widespread use, particularly because the user can control the morph by specifying feature correspondences. Morphing polyhedral objects would be advantageous in several applications; however, it lacks easy user interaction.
In this paper, we present techniques for morphing between simple polyhedra, taking into account user defined features. This is achieved in three steps: First, the vertex-edge graph of the polyhedron is embedded on a sphere. Second, these embeddings are warped so that corresponding vertices coincide. Third, the warped embeddings are merged. Thus, corresponding feature vertices are fused in the merged model and, consequently, features are preserved in any transformation. Additionally, transition control - a concept from image morphing - is introduced in polyhedral morphing, yielding more freedom in the generation of animations. Even more importantly, this opens up new ways of modeling polyhedral objects.

9:30 Ziv Bar-Joseph
Computer Science Dept.
The Hebrew University Texture Movies: Statistical Learning of Time-Varying Textures We present an algorithm for synthesizing new instances of a time-varying texture, given an example of such a texture as input. This algorithm makes it possible to synthesize movie clips of dynamic phenomena such as waterfalls, fire flames, a school of jellyfish, turbulent clouds, an erupting volcano, and a crowd of people. A one-dimensional variant of our algorithm is able to synthesize various sound textures, such as traffic, water sounds, etc. The input time-varying texture is treated as a sample of a 3D signal generated by a stochastic process. We first construct a tree representing a hierarchical multi-scale transform of the signal using a combination of Daubechies wavelets and steerable pyramid transforms. >From this tree, new random trees are generated by learning and sampling the conditional probabilities of the paths in the original tree. Transformation of these random trees back to signals results in new texture movies and/or sound textures.
Joint work with Ran El-Yaniv (Technion), Dani Lischinski (HUJI), Michael Werman (HUJI)

10:00 Jacky Romano
Silicon Graphics Inc. Israel Silicon Graphics 320 Visual Workstation - Architecture and Applications The session will present the Unified Memory Architecture of the Silicon Graphics 320, and its usage in various applications, such as:
* Multi Pass rendering.
* Video to Texture
* Large Images display
* Volume Rendering

10:30 COFFEE BREAK

11:15 Leo Joskowicz
Computer Science Dept.
The Hebrew University The Configuration Space Method for Kinematic Tolerance Analysis We present an overview of a new kinematic tolerance analysis algorithm for mechanical systems with parametric part tolerances. The algorithm constructs a variation model for the system, derives worst-case bounds on the variation, and helps designers find unexpected failure modes, such as jamming and blocking. The variation model is a generalization of the configuration space representation of the part contacts in the nominal system. The algorithm handles general planar systems of curved parts with contact changes, including open and closed kinematic chains. It constructs a variation model for each interacting pair of parts and then derives the overall system variation at a given configuration by composing the pairwise variation models via sensitivity analysis and linear programming. Our implementation analyzes systems with up to 100 parameters in under a minute on a workstation.
Joint work with Elisha Sacks, Computer Science Dept, Purdue University, USA.

11:45 Vitaly Surazhsky
Computer Science Dept. Technion Morphing Planar Triangulations Morphing, also known as metamorphosis, is the gradual and continuous transformation of one shape into another. The morphing problem has been investigated in many contexts, e.g., morphing of two-dimensional images, polygons, polylines and freeform curves. Triangulation is frequently used in computer graphics as a parameterization for surfaces and images. Two planar triangulations of corresponding point sets are said to be compatible if they are topologically equivalent. This work describes methods for morphing compatible planar triangulations. Triangulations with fixed boundary are considered as a particular case.
Our method is based on the representation of a planar triangulation as a matrix derived from the triangulation topology. The matrix is constructed using barycentric coordinates of the vertices relative to their neighbors. The choice of such a matrix has many degrees of freedom. Morphing the triangulations corresponds to interpolations between matrices, which may be performed in different ways. This work shows that the method always yields a valid morph, namely all intermediate graphs are valid planar triangulations. In the context of the morphing problem, this means that the morph is self-intersection free. This gives an exact solution to a problem which has been solved up to now only by a variety of heuretics. Two techniques for interpolation between matrices are proposed. The first one uses convex combinations of the matrices. The second technique performs a more sophisticated ``interpolation'' between matrices, in which each row of the matrices is treated separately, in a ``local'' manner. We investigate properties of the morphs generated using matrices.
We study the trajectories traveled by the vertices during the morph and reducibility to a linear-trajectory morph. In addition, we show that the method can be extended to deal with constrained morphs. A morph through a fixed intermediate triangulation can be derived using this method. It is also demonstrated how to obtain a morph with a natural evolution of triangle areas.
Joint work with Craig Gotsman.

12:15 Ariel Manor
Mechanical Eng. Dept.
Technion Recognition of 3D Models through Integration of Processing Techniques in Reverse Engineering Laser scanners are commonly used in Reverse Engineering since they can digitize 3D objects fast and very accurately relative to other systems. However, a laser scanner system provides a huge number of sample points that require intensive processing in order to reconstruct the surface of the object. When data reduction algorithms are applied to the data, the smaller data set becomes irregular and scattered and cannot be handled by ordinary parametric surface approximation methods.
An appropriate parameterization technique for unordered points is the base surface parameterization method, in which each sample point is projected onto a parametric surface to determine its parameter values. The base surface is a rough approximation of the surface. In this work, the base surface is generated from a 3D boundary curve. This guiding curve can be extracted from the data by integrating additional information derived from an image of the object. It appears that combining 3D scanned data with 2D intensity images can improve the reconstruction process. 3D scanned data provides explicit 3D coordinates, whereas an intensity image offers other 2D characteristic information such as texture and boundary. For smooth freeform surfaces that generate smooth 2D silhouettes, characteristic curves of the surface, such as boundaries, can be derived from its image using image processing techniques. In this research, a new adaptive parametric surface reconstruction method is proposed. The method integrates information from a CCD camera image with digitized points from a laser scanner. The method is based on the following stages: (a) Extracting a 3D boundary curve of the surface by matching an intensity image with the scanned data using image processing techniques; (b) Generating a parametric base surface which is guided by the extracted boundary curve; (c) Parameterizing the 3D sample points using the base surface; and (d) Fitting a B-spline surface by applying a least squares (LSQ) approximation method with boundary constraints. The feasibility of the surface reconstruction algorithm is demonstrated by several examples of sculptured freeform surfaces.
Joint work with Anath Fischer

Time	Speaker	Title	Abstract
8:30		REFRESHMENTS
9:00	Marc Alexa Computer Science Dept. Darmstadt Univ. of Technology, Germany	Morphing Polyhedral Shapes with Scattered Features	Metamorphosis techniques for images and volume models have gained widespread use, particularly because the user can control the morph by specifying feature correspondences. Morphing polyhedral objects would be advantageous in several applications; however, it lacks easy user interaction. In this paper, we present techniques for morphing between simple polyhedra, taking into account user defined features. This is achieved in three steps: First, the vertex-edge graph of the polyhedron is embedded on a sphere. Second, these embeddings are warped so that corresponding vertices coincide. Third, the warped embeddings are merged. Thus, corresponding feature vertices are fused in the merged model and, consequently, features are preserved in any transformation. Additionally, transition control - a concept from image morphing - is introduced in polyhedral morphing, yielding more freedom in the generation of animations. Even more importantly, this opens up new ways of modeling polyhedral objects.
9:30	Ziv Bar-Joseph Computer Science Dept. The Hebrew University	Texture Movies: Statistical Learning of Time-Varying Textures	We present an algorithm for synthesizing new instances of a time-varying texture, given an example of such a texture as input. This algorithm makes it possible to synthesize movie clips of dynamic phenomena such as waterfalls, fire flames, a school of jellyfish, turbulent clouds, an erupting volcano, and a crowd of people. A one-dimensional variant of our algorithm is able to synthesize various sound textures, such as traffic, water sounds, etc. The input time-varying texture is treated as a sample of a 3D signal generated by a stochastic process. We first construct a tree representing a hierarchical multi-scale transform of the signal using a combination of Daubechies wavelets and steerable pyramid transforms. >From this tree, new random trees are generated by learning and sampling the conditional probabilities of the paths in the original tree. Transformation of these random trees back to signals results in new texture movies and/or sound textures. Joint work with Ran El-Yaniv (Technion), Dani Lischinski (HUJI), Michael Werman (HUJI)
10:00	Jacky Romano Silicon Graphics Inc. Israel	Silicon Graphics 320 Visual Workstation - Architecture and Applications	The session will present the Unified Memory Architecture of the Silicon Graphics 320, and its usage in various applications, such as: * Multi Pass rendering. * Video to Texture * Large Images display * Volume Rendering
10:30		COFFEE BREAK
11:15	Leo Joskowicz Computer Science Dept. The Hebrew University	The Configuration Space Method for Kinematic Tolerance Analysis	We present an overview of a new kinematic tolerance analysis algorithm for mechanical systems with parametric part tolerances. The algorithm constructs a variation model for the system, derives worst-case bounds on the variation, and helps designers find unexpected failure modes, such as jamming and blocking. The variation model is a generalization of the configuration space representation of the part contacts in the nominal system. The algorithm handles general planar systems of curved parts with contact changes, including open and closed kinematic chains. It constructs a variation model for each interacting pair of parts and then derives the overall system variation at a given configuration by composing the pairwise variation models via sensitivity analysis and linear programming. Our implementation analyzes systems with up to 100 parameters in under a minute on a workstation. Joint work with Elisha Sacks, Computer Science Dept, Purdue University, USA.
11:45	Vitaly Surazhsky Computer Science Dept. Technion	Morphing Planar Triangulations	Morphing, also known as metamorphosis, is the gradual and continuous transformation of one shape into another. The morphing problem has been investigated in many contexts, e.g., morphing of two-dimensional images, polygons, polylines and freeform curves. Triangulation is frequently used in computer graphics as a parameterization for surfaces and images. Two planar triangulations of corresponding point sets are said to be compatible if they are topologically equivalent. This work describes methods for morphing compatible planar triangulations. Triangulations with fixed boundary are considered as a particular case. Our method is based on the representation of a planar triangulation as a matrix derived from the triangulation topology. The matrix is constructed using barycentric coordinates of the vertices relative to their neighbors. The choice of such a matrix has many degrees of freedom. Morphing the triangulations corresponds to interpolations between matrices, which may be performed in different ways. This work shows that the method always yields a valid morph, namely all intermediate graphs are valid planar triangulations. In the context of the morphing problem, this means that the morph is self-intersection free. This gives an exact solution to a problem which has been solved up to now only by a variety of heuretics. Two techniques for interpolation between matrices are proposed. The first one uses convex combinations of the matrices. The second technique performs a more sophisticated ``interpolation'' between matrices, in which each row of the matrices is treated separately, in a ``local'' manner. We investigate properties of the morphs generated using matrices. We study the trajectories traveled by the vertices during the morph and reducibility to a linear-trajectory morph. In addition, we show that the method can be extended to deal with constrained morphs. A morph through a fixed intermediate triangulation can be derived using this method. It is also demonstrated how to obtain a morph with a natural evolution of triangle areas. Joint work with Craig Gotsman.
12:15	Ariel Manor Mechanical Eng. Dept. Technion	Recognition of 3D Models through Integration of Processing Techniques in Reverse Engineering	Laser scanners are commonly used in Reverse Engineering since they can digitize 3D objects fast and very accurately relative to other systems. However, a laser scanner system provides a huge number of sample points that require intensive processing in order to reconstruct the surface of the object. When data reduction algorithms are applied to the data, the smaller data set becomes irregular and scattered and cannot be handled by ordinary parametric surface approximation methods. An appropriate parameterization technique for unordered points is the base surface parameterization method, in which each sample point is projected onto a parametric surface to determine its parameter values. The base surface is a rough approximation of the surface. In this work, the base surface is generated from a 3D boundary curve. This guiding curve can be extracted from the data by integrating additional information derived from an image of the object. It appears that combining 3D scanned data with 2D intensity images can improve the reconstruction process. 3D scanned data provides explicit 3D coordinates, whereas an intensity image offers other 2D characteristic information such as texture and boundary. For smooth freeform surfaces that generate smooth 2D silhouettes, characteristic curves of the surface, such as boundaries, can be derived from its image using image processing techniques. In this research, a new adaptive parametric surface reconstruction method is proposed. The method integrates information from a CCD camera image with digitized points from a laser scanner. The method is based on the following stages: (a) Extracting a 3D boundary curve of the surface by matching an intensity image with the scanned data using image processing techniques; (b) Generating a parametric base surface which is guided by the extracted boundary curve; (c) Parameterizing the 3D sample points using the base surface; and (d) Fitting a B-spline surface by applying a least squares (LSQ) approximation method with boundary constraints. The feasibility of the surface reconstruction algorithm is demonstrated by several examples of sculptured freeform surfaces. Joint work with Anath Fischer