Catálogo de publicaciones - libros

Traditional simplification algorithms operate either in the main memory or in the disk space. In this paper, we propose a novel storage independent polygonal mesh simplification system (SIPMSS). The new system offers an advanced memory utilization scheme and efficient storage-independent primitives that achieve constant main memory footprint and excellent runtime efficiency. It not only provides a very flexible platform for both in-core and out-of-core simplification algorithms but also permits very efficient integration of various types of simplification algorithms. The results presented in the section of experiments further approve its effectiveness and usefulness for both in-core and out-of-core of simplification.

In order to get better rendering effect, we proposed a wavelet transform based image enhancement algorithm for real time multi-resolution texture mapping in this paper. In this algorithm, we proposed a concept of weighted pyramid. Texture images to be synthesized are weighted enhanced by wavelet transform. Comparing with original ones, the edges in this image is much clearer and the contrast of this image is more evident. And when dealing with the maps with their particular characteristics, we also proposed some method for algorithm improvement. By using this method, current multi-resolution texture mapping method will be much more improved. It can give us a more realistic and efficient rendering effect by using lower resolution image to replace higher resolution one.

In this paper, stereoscopic image visualization for teleoperated robots is studied. The use of stereo vision systems shows some difficulties: image disparity is the main problem. It affects the correct image fusion process by the brain. Disparity curve calculations are shown and a method to calibrate stereoscopic images for telerobotics is explained. Experiments have been carried out to obtain the limits for a proper image fusion. The use of stereoscopic image has been assessed to execute teleoperated robot guidance tasks. The performance of stereoscopic and monoscopic image is compared. Stereoscopic images have shown a guaranteed depth perception.

This paper describes various techniques of accessing objects hidden behind obstacles in virtual environments. We show some of well known interaction techniques that are suitable for usage in this case. Next, we introduce new Head Manipulator method and several aids that allow a user to change the visibility of objects. We propose several combinations of interaction techniques and discuss their suitability for the given task. Finally, we describe our experiments when testing these methods and then we conclude with the results from the experiments and from users’ observations.

This paper illustrates a novel registration method based on robust estimation of trifocal tensors using point and line correspondences synthetically. The proposed method distinguishes itself in following ways: Firstly, besides feature points, line segments are also used to calculate the needed trifocal tensors. The registration process can still be achieved even under poorly textured scenes. Secondly, to estimate trifocal tensors precisely, we use PROSAC instead of RANSAC algorithm to remove outliers. While improving the accuracy of our system, PROSAC also reduces the computation complexity to a large degree.

This paper investigates methods of describing two-dimensional and three-dimensional scenes using eXtensible Mark-up Language (XML). It also investigates the initial development of a non-commercial object-based scene description language, for the effective capture of two- and three-dimensional scenes. Design science research techniques were applied to determine the ideal factors involved in effective scene capture. The solution facilitates the description of 2D and 3D environments via interpretation of object relationships; the implementation of the inheritance, functionality, interactions and behaviour.

High-quality mesh provides good rendering effects of virtual reality systems. In this paper a solid technique for generating triangle meshes optimized for real-time rendering in virtual environments is proposed. Modified weighted Voronoi barycenter relaxation is used to improve the mesh geometry. A simulated annealing algorithm is presented for optimizing vertex connec tivity. The shape deviations are effectively prevented, and the boundaries and features are well preserved. The advantage and robustness of this technique are verified by many examples.

Terrain rendering is essential to represent virtual environment in interactive applications. We have explored voxel-based terrain rendering using ray casting method. Real-time applications such as computer games and simulations require interactive fly-through or walk-through on the terrain. However, since previous approaches usually require a tremendous amount of CPU computation for view-frustum culling and level-of-detail selection, they are not suitable for those applications. We propose an accelerated terrain rendering method using GPU, which exploits a novel data structure, named , for the acceleration of the voxel-based terrain rendering. In addition, we expanded the conventional PARC (Polygon Assisted Ray Casing) algorithm by applying our method to accelerate rendering performance of GPU- based terrain rendering method.

In this paper we present an illumination model for rendering realistic stained glass which simulates the phenomenon of stained glass in real world. The optics for stained glass involves three basic physical mechanisms: diffuse light and specular highlight, Fresnel refraction, volume absorption; these dominate the appearance of stained glass. In this paper, we combine them to render realistic stained glass. Our approach avoids performing any expensive geometric tests, and involves no pre-computation; both of which are common features in previous work. The rendered stained glass images achieve excellent realism.

Visual hull is intersection of cones made by back-projections of reference images. We introduce a real-time rendering method of jaggy-free visual hull on programmable graphics hardware. Visual hull can be rendered quickly by using texture mapping approach. However, jagged artifacts are present near the edges of visual hull faces. In this paper, we solve the problem by using silhouette information. Our implementation demonstrates high-quality rendering results in real-time. Time complexity of our algorithm is , where is the number of reference images. Thus, the examples in this paper are rendered over one hundred of frames per second without jaggies.