Catálogo de publicaciones - libros

We present a simple closed form solution for smooth quaternion interpolation problem. In contrast to other methods, our approach does not rely on cubic B-spline blending functions which require solution of nonlinear tridiagonal system. Instead, we propose using interpolatory (cardinal) basis. Our method outperforms all alternatives and, being explicit, is absolutely stable.

Fluid animations are becoming a standard tool for computer animators. Simulation of turbulent gases, running water, eroded surfaces, or splashing waves are common, but still demanding because they are usually calculated in a voxel space. This brings new requirements to the tools that are used for such animations. The data structures are enormous but providing a good space and time coherency. We present a compression scheme that can be used for storing, accessing, and viewing such animations interactively. Key-frames are compressed by the RLE algorithm and in-betweens as difference frames. To display the scene we convert the level of water and the terrain surface to triangle meshes by the marching cubes algorithm. With this lossless technique we reach compression factor up to 1:100. Scenes can be decompressed fast, can be displayed, and manipulated interactively.

This paper presents a simple, effective method for constructing elbow and knee joints between rigid NURBS-based human body segments for animation. The purpose is to automatically and efficiently manipulate NURBS joint patches to provide smooth, realistic joint action. The production of fine surface detail is also illustrated by creating wrinkles in the joint acute angle, also in a simple, effective manner. By automating such processes the need to manually add surface detail is reduced; animator input and animation time is decreased.

This paper presents a method for motion recovery from monocular images containing human motions. Image processing techniques, such as spatial filter, linear prediction, cross correlation, least square matching etc, are applied to extract feature points from 2D human figures with or without markers. A 3D skeleton human model is adopted with encoded angular constraints. Energy Function is defined to represent the residuals between extracted feature points and the corresponding points resulted from projecting the human model to the projection plane. Then a procedure for motion recovery is developed, which makes it feasible to generate realistic human animations

In the paper basic problems related to audiovisual synthesis for Polish are described. Issues of the lips synchronization in two- and three-dimensional animation are presented. As results of the work a so called “chatting head” have been prepared which can say all words typed at the computer keyboard.

This paper deals with representations of artefacts used during computer-aided design process. Two representations are proposed: external, based on XML format and internal to the application, object-oriented. A graphical visualization of designed model on skeletal structure example is also presented.

The paper proposes an extension of fractal generation with the use of chaos games to multidimensional spaces. Brightness of color is used to simulate additional higher dimensions. Examples of the generated fractals are shown.

The paper presents a new method of creating contour lines based on irregularly spaced data. The analysis has been focused on contour plots as an useful tool of visualization and real-time verification of the postprocessing stage in multigrid computational environment. A new effective contouring concept has been introduced and developed to satisfy this objective. The presented approach is based on the OpenGL.

The main problem of virtual art gallery visualization is achieved compromise between a high cost of calculations (a requirement of real-time rendering), and a level of the image realism (the global illumination is needed). The paper analyses various meshing techniques for radiosity method. For one of them (i.e. Discontinuity Meshing) the original method of accurate updating radiosities for new vertices is presented.

This work has to do with Optical Character Recognition in the context of degraded characters, and more particularly in the context of cut characters. It appears like a complement to a more general problem and investigates a field not often seen in literature. The problem is first divided into two sub-problems: first, modular networks, separating the detection of cut characters and their classification are proposed. Then, the classification part is completed with information coming from the prolongation of cut characters.