Catálogo de publicaciones - libros

The paper introduces a new, computationally inexpensive, real-time method of simulation of clouds for dynamic particle-based cloud rendering. The method covers such phenomena as atmospheric fronts, stable and unstable balance, thunderstorms, simplified simulation of middle and high clouds, correlation between wind and clouds and even tornadoes. The simulation is based on several simple numeric global parameters, like average temperature and average amount of clouds, which determine the local weather. It is also capable of generating multiple forms of clouds (from through ), which may fluently change during the simulation.

Most existing methods for determining localization of the image feature point are still inefficient in terms of the precision. In the paper, we propose a new algorithm for feature point detection based on the combined intensity variation status along the adaptive principal direction of the corner. Firstly, we detect principal orientation of each pixel, instead of calculating the gradients along the horizontal and vertical axes. And then we observe the intensity variations of the pixel along the adaptive principal axes and its tangent one respectively. When the combined variation status is classified into several specific types, it can be used to determine whether a pixel is a corner point or not. In addition to corner detection, it is also possible to use our proposed algorithm to detect the edges, isolated point and plain regions of a natural image. Experimental results on synthetic and natural scene images have shown that the proposed algorithm can successfully detect any kind of the feature points with good accuracy of localization.

In this paper we present a method for fast energy minimization of virtual garments. Our method is based upon the idea of multi-resolution particle system. When garments are approximately positioned around a virtual character, their spring energy may be high, which will cause instability or at least long execution time of the simulation. An energy minimization algorithm is needed; if a fixed resolution is used, it will require many iterations to reduce its energy. Even though the complexity of each iteration is O(n), with a high resolution mass-spring system, this minimization process can take a whole day. The hierarchical method presented in this paper is used to reduce significantly the execution time of the minimization process. The garments are firstly discretized in several resolutions. Once the lowest resolution particles system is minimized (in a short time), a higher resolution model is derived, then minimized. The procedure is iterated up to the highest resolution. But at this stage, the energy to minimize is already much lower so that minimization takes a reasonable time.

Human motion analysis has received a great attention from researchers in the last decade due to its potential use in different applications. We propose a new approach to extract human joints (vertex positions) using a model-based method. Motion templates describing the motion of the joints as derived by gait analysis, are parametrized using the elliptic Fourier descriptors. The heel strike data is exploited to reduce the dimensionality of the parametric models. People walk normal to the viewing plane, as major gait information is available in a sagittal view. The ankle, knee and hip joints are successfully extracted with high accuracy for indoor and outdoor data. In this way, we have established a baseline analysis which can be deployed in recognition, marker-less analysis and other areas. The experimental results confirmed the robustness of the proposed method to recognize walking subjects with a correct classification rate of %92.

Non-Photorealistic Rendering (NPR) can offer increased concentration and familiarity to the watcher. For this reason, many media such as movies, games, and commercials currently use the NPR method to deliver information. In this paper, we suggest an interactive system for video cartooning based on the mean-shift segmentation of image and video. In order to solve the problems of time complexity and memory allocation, the conventional problems of video mean-shift segmentation, this paper proposes several techniques such as foreground object based segmentation and sequential segmentation. We also propose the interactive correction technique to get enhanced results. For more cartoonic representation, we used spline curve approximation of segment boundaries in the final rendering results. With our method, we can easily create cartoon rendering output using the video streams such like home video, which can be obtained easily in our daily life.

Virtual Reality techniques were used to develop educational didactic models in the area of Civil Engineering. The visualization of the distinct physical steps of a construction is shown in the virtual applications here presented. The developed models bring new perspectives in the teaching activity as a support to expose new curricular programmes or complex sequence construction. In order to obtain models, which would be able to visually simulate the geometric evolution of the construction activity, techniques of geometric modelling and virtual reality were used. The models make it possible to view the physical evolution of the work, to follow the planned construction sequence and to visualize details of the form of every component of the works. They also support the study of the type and method of operation of the equipment necessary for these construction procedures. These models have been used to distinct advantage as educational aids in first-degree courses in Civil Engineering.

In this paper we present an agent that can analyse certain human full-body movements in order to respond in an expressive manner with copying behaviour. Our work focuses on the analysis of human full-body movement for animating a virtual agent, called Greta, able to perceive and interpret users’ expressivity and to respond properly. Our system takes in input video data related to a dancer moving in the space. Analysis of video data and automatic extraction of motion cues is done in EyesWeb. We consider the amplitude and speed of movement. Then, to generate the animation for our agent, we need to map the motion cues on the corresponding expressivity parameters of the agent. We also present a behaviour markup language for virtual agents to define the values of expressivity parameters on gestures.

This paper presents a new low-dimensional face representation using the proposed eigenspaces transformation. The proposed algorithm is based on face images which is acquired with illumination conditions. We define face images as a non-illumination class and illumination classes from light source conditions and derive the linear transformation function in a low-dimensional eigenspace between a non-illumination class and illumination classes. The proposed illumination compensation algorithm is composed of two steps. In the optimal projection space which is obtained from the DirectLDA algorithm, we first select the illumination class for a given image and then we generate a non-illuminated image by using eigenspace transformation of the illuminated class. We provide experimental results to demonstrate the performance of the proposed algorithm with varying parameters of proposed algorithm.

Segmenting garments from humanoid meshes or point clouds is a challenging problem with applications in the textile industry and in model based motion capturing. In this work we present a physically based template-matching technique for the automatic extraction of garment dimensions from 3D meshes or point clouds of dressed humans. The successfull identification of garment dimensions also allows the semantic segmentation of the mesh into naked and dressed parts.

This paper presents a method for 3D face modeling from a set of 2D images. This method is based on deformation of a pre-defined generic polygonal face mesh to the specific face of the person, presented on several images taken from different views. This deformation is based upon matching points and silhouette curves. To provide high accuracy we have chosen an interactive way of getting information from the image. Firstly the user defines several feature points on the images. Then the algorithm adapts the generic model to the real face presented on the image by matching the projections of the generic characteristic points to those of the images. Finally, the reconstruction is improved by matching the silhouettes of the deformed model to those of the images.