Catálogo de publicaciones - libros

In this paper, we present a system for completing the blank hole in an image list or a video sequence, which can be used in movie-making industry to produce some special montage effect. To achieve this, we apply a 3D coordinate to depict the video clip. As the time information representing different frames is considered, the available information is enriched than the single image. Our method takes the global character into account for avoiding visual inconsistencies. We view the video completion problem as a labeling problem and apply the belief propagation algorithm to minimize the energy defined on the label graph. By measuring the similarity between source and target patches, we could discover the appropriate patch to fill the hole. Furthermore, we introduce some techniques to speedup the belief propagation algorithm for gaining better performance efficiency. Examples demonstrate the effectiveness of our approach and the results are encouraging.

With the increasing popularity of digital cameras, image editing tools for amateur users are needed. Patrick Pérez and his colleagues generalized that image editing tasks concern either global changes or local changes confined to a selection[1]. Here we focus on local editing of images’ color—object-based image recoloring.

In this paper, we introduce a framework for recoloring destination region (desired object) with desired color, which integrates alpha matting and compositing and color transformation algorithms to provide an object-based recoloring tool for common users. There are three steps in the framework: matte extraction, color transformation and alpha compositing. We present a fast color histogram specification as a general method for example-based color transformation. In some cases, our framework to the recoloring problem are quite simple but works well. We thus feel that the primary contribution of this paper may lie in the organic integration of algorithms totally.

Recent advances in single view reconstruction (SVR) have been in modeling 3D curved surfaces and automation. We extend the SVR along the first direction in several ways: (i) We study the reconstruction of curved surfaces under perspective projection rather than orthographic projection. (ii) We illustrate that the horizontally flipped image of the reflective symmetric objects is equivalent to an image photographed from another position under the simplified perspective projection. So the traditional structure from motion techniques can be used to recover the 3D feature points on them. They can be used to create the planar model of the object and be used as position constraints to generate curved surfaces. (iii) New linear constraints, such as perspective projection, are exploited to the linearly constrained quadratic optimization to finding the smoothest surface. We demonstrate these advances in reconstructing real symmetric curved surfaces and objects.

This paper researches the method of constructing 3D surface from planar contours. We base our work on distance field function method and mainly concentrate on how to simply and uniformly solve the problems of isosurface generation caused by non-manifold surface. Our work includes three main steps: grid adjustment analysis, volume construction and surface construction. In the first step, we present a new method to process non-manifold contour by adaptively adjusting grid size. In the volume construction and the surface construction steps, we use classic distance field function and Marching Cube method respectively. The experiment shows that our algorithm has more realistic results in constructing 3D surface from planar contour.

In this paper we present a new method for the model of interpolation surfaces by the blending of polar coordinates and Cartesian coordinate. A trajectory curve is constructed by circular trigonometric Hermite interpolation spline (CTHIS) and a profile curve is presented by -continuous B-spline like interpolation spline (BSLIS). A piecewise interpolation spline surface is incorporated by the blending of CTHIS and BSLIS. In addition, scaling functions have been introduced to improve the flexibility of the model of the interpolation surfaces. On the basis of these results, some examples are given to show how the method is used to model some interesting surfaces.

Two contributions on 3D implicit surface reconstruction from scattered points are presented in this paper. Firstly, least square radial basis functions (LS RBF) are deduced from the conventional RBF formulations, which makes it possible to use fewer centers when reconstruction. Then we use orthogonal least square (OLS) method to select significant centers from large and dense point data sets. From the selected centers, an implicit continuous function is constructed efficiently. This scheme can overcome the problem of numerical ill-conditioning of coefficient matrix and over-fitting. Experimental results show that our two methods are efficient and highly satisfactory in perception and quantification.

This paper proposes a new interpolatory subdivision scheme for triangular meshes that produces continuous limit surfaces for both regular and irregular settings. The limit surfaces also have bounded curvature, which leads to improved quality surfaces. The eigen-structure analysis demonstrates the smoothness of the limit surfaces. According to the new scheme, the approach for progressive transmission of meshes is presented. Finally, results of refined models with the new scheme are shown. In most cases, the new scheme generates more pleasure surfaces than the traditional modified butterfly scheme, especially near the irregular settings.

Based on the objective function defined by the approximation of the curvature variation formula of the curve, a new method for constructing composite optimized geometric Hermite (COH) curves is presented in this paper. The new method can deal with some cases in which neither of the existing methods those are based on minimum curvature variation or minimum strain energy can get pleasing shape. The comparison of the new method with the existing methods are given, which shows that none of the new method and the existing ones can deal with all the cases well. The experiments show that combination of the new method with the existing methods can achieve a good result in all cases.

PCA is a useful tool for pose normalization in 3D model retrieval, in this paper, we analyze its basic principle, point out its shortcoming and give our solution. Experiments show that our methods can enhance the robustness of PCA and align 3D models well.

This paper presents a system for crowd evacuation in emergency situation based on dynamics model to offer a base platform for further researches. Starting with the implementation of base functions, our focus is on the stability and expandable of the platform to offer new functions easily according to our needs latter. To improve the independence of the module, the function into layers and dividing the data are separated into blocks. To reach efficient system implementation, the 3D building is translated into a 2D graphics by turning the map into a group of nodes. Furthermore, the element called node plug is used to enhance the expansibility of the system. Experiments are carried out to analyze the crowd’s evacuation efficiency in a given building. The impact caused by mass behavior, the structure of the building and the number of people inside are also construed qualitatively in the experiments.