Catálogo de publicaciones - libros

This paper presents an improved color mood blending between images via fuzzy relationship. We take into consideration the weighted influences of the source as well as the target image. Our algorithm automatically calculates the weights according to the fuzzy relations of images with Gaussian Membership Function, derived from both the statistical features of the source and target image. As the experimental results shown, the visual appearance of the resulting image is more natural and vivid. Our algorithm can offer users another selection for perfecting their work. It has four advantages. First, it is a general approach where previous methods are special cases of our method. Second, it produces a new style and feature. Third, the quality of the resultant image is visually plausible. Finally, it is simple and efficient, with no need to generate swatches.

Alzheimer’s disease is the most common cause of dementia, yet hard to diagnose precisely without invasive techniques, particularly at the onset of the disease. This work approaches image analysis and classification of synthetic multispectral images composed by diffusion-weighted magnetic resonance (MR) cerebral images for the evaluation of cerebrospinal fluid area and its correlation with the advance of Alzheimer’s disease. The MR images were acquired from an image system by a clinical 1.5 T tomographer. The classification methods are based on multilayer perceptrons, polynomial nets and Kohonen LVQ classifiers. The classification results are used to improve the usual analysis of the apparent diffusion coefficient map.

To overcome the ill-posed problem of image reconstruction with noisy detected data in PET reconstruction, Bayesian reconstruction or maximum a posteriori (MAP) method has its superiority over others with regard to image quality and convergence. Based on Markov Random Fields (MRF) and Bayesian reconstruction theory, quadratic membrane (QM) prior and quadratic plate (QP) prior function differently for different objective surfaces with different properties. It is reasonable to believe that a hybrid prior which combines the two quadratic prior can work better than just using one prior alone. In this paper, a MRF quadratic hybrid prior multi-order model is proposed. A threshold estimation method based on statistical classification is devised to facilitate a selectively utilization of QM prior, QP prior in the quadratic hybrid multi-order (QHM) prior. Application of the proposed QHM prior in PET reconstruction with joint estimation algorithm is also given. Visional and quantitative comparisons of the results of experiments prove the new hybrid prior’s good performance in lowering noise effect and preserving edges for PET reconstruction.

We focus on improving the effectiveness of content-based 3D shape retrieval. Motivated by retrieval performance of several individual 3D model feature vectors, we propose a novel method, called prior knowledge based automatic weighted combination, to improve the retrieval effectiveness. The method dynamically determines the weighting scheme for different feature vectors based on the prior knowledge. The experimental results show that the proposed method provides significant improvements on retrieval effectiveness of 3D shape search with several measures on a standard 3D database. Compared with two existing combination methods, the prior knowledge weighted combination technique has gained better retrieval effectiveness.

In our previous work [21], we have shown that by means of a model-based approach, relightable free-viewpoint videos of human actors can be reconstructed from only a handful of multi-view video streams recorded under calibrated illumination. To achieve this purpose, we employ a marker-free motion capture approach to measure dynamic human scene geometry. Reflectance samples for each surface point are captured by exploiting the fact that, due to the person’s motion, each surface location is, over time, exposed to the acquisition sensors under varying orientations. Although this is the first setup of its kind to measure surface reflectance from footage of arbitrary human performances, our approach may lead to a biased sampling of surface reflectance since each surface point is only seen under a limited number of half-vector directions. We thus propose in this paper a novel algorithm that reduces the bias in BRDF estimates of a single surface point by cleverly taking into account reflectance samples from other surface locations made of similar material. We demonstrate the improvements achieved with this spatio-temporal reflectance sharing approach both visually and quantitatively.

As the sampling data is getting tremendous, more than one sampling points will project into a pixel. This makes point based rendering (PBR) popular. For PBR, the main steps that prominently affect the rendering result are hierarchical data structure, LOD selection method and rendering primitives (triangle, point, surfel [1]). In this paper, we generate a hierarchical structure with tight-octree, and store the vertex and bounding box information for each level. Then we propose a new method to do LOD selection based on the distance between the model and the viewer and the pre-calculated bounding box information. We have tested different polygonal models with million vertices on our system and the results demonstrate that the method is interactive in real time.

We present a novel FPGA-accelerated architecture for fast collision detection among rigid bodies. This paper describes the design of the hardware architecture for several primitive intersection testing components implemented on a multi-FPGA Xilinx Virtex-II prototyping system. We focus on the acceleration of ray-triangle intersection operation which is the one of the most important operations in various applications such as collision detection and ray tracing.

Our implementation result is a hardware-accelerated ray-triangle intersection engine that is capable of out-performing a 2.8 GHz Xeon processor, running a well-known high performance software ray-triangle intersection algorithm, by up to a factor of seventy. In addition, we demonstrate that the proposed approach could prove to be faster than current GPU-based algorithms as well as CPU based algorithms for ray-triangle intersection.

Nowadays large volumes of data with high dimensionality are being generated in many fields. ClusterTree is a new indexing approach representing clusters generated by any existing clustering approach. It supports effective and efficient image retrieval. Lots of clustering algorithms have been developed, and in most of them some parameters should be determined by hand. The authors propose a new ClusterTree structure, which based on the improved CLIQUE and avoids any parameters defined by user. Using multi-resolution property of wavelet transforms, the proposed approach can cluster at different resolution and remain the relation between these clusters to construct hierarchical index. The results of the application confirm that the ClusterTree is very applicable and efficient.

Due to the widely application of 3D models, the techniques of content-based 3D shape retrieval become necessary. In this paper, a modified Principal Component Analysis (PCA) method for model normalization is introduced at first, and each model is projected in 6 different viewpoints. Secondly, a new adjacent angle distance Fouriers (AADF) descriptor is presented, which captures more precise contour feature of black-white images. Finally, based on modified PCA method, a novel composite 3D shape descriptor MADE is proposed by concatenating AADF, Tchebichef and D-buffer descriptors. Experimental results on the criterion of 3D model database PSB show that the proposed descriptor MADE has gained the best retrieval effectiveness compared with three single descriptors and two composite descriptors LFD and DESIRE.

Chinese calligraphy is a precious Chinese art. It is pictographic and each calligraphist has his own writing style. Even people whose native language is Chinese will have difficulties in writing a demanded beautiful calligraphy style, not to say people who know little about Chinese calligraphy. In order to help people enjoy the art of calligraphy writing and find out the process how it was written, we implement a new approach to animate its writing process by 3D visualization method. In the paper two novel algorithms are also presented to extract the intrinsic feature which is needed to rebuild the writing process: 1) extract strokes order from an offline Chinese calligraphic handwriting; 2) estimate varied stroke’s thickness. Finally, experiment result is given to demonstrate the application.