Catálogo de publicaciones - libros

A robust, adaptive system for reconstructing 3D human motion from monocular video is presented. Our system takes a model based approach. To save computation time, we fit the skeleton instead of full body model to the silhouette. End sites positions identified from the silhouettes serve as an extra constraint. The alleviation of computational burden then makes the use of simulated annealing practical to get rid of local minima. The identifiable end sites also serve as the criterion to judge the reconstruction reliability of single frames. According to different reliabilities, the video is segmented into sections, which are reconstructed using different strategies. Our system is robust, getting rid of error accumulation in tracking, and adaptive, being able to tell the user when and where more information is needed.

Wavelet scalable video coding (WSVC) can be used efficiently for scalable mesh compression for virtual reality applications over IP networks. This paper proposes efficient methods to improve the coding efficiency of WSVC through the minimization of average temporal distance (ATD) for motion estimation. The ATD is closely related to the peak signal-to-noise ratio (PSNR) performance of WSVC which is based on motion compensated temporal filtering. Simulation results show that the proposed methods with minimum ATD give about 1 to 2dB PSNR improvements compared to the conventional WSVC.

Most available gaze tracking systems based on corneal reflection have low tolerance for head movements. In this paper, an eye-gaze tracking method, which can make this system be adaptive to external lighting changes and to the user’s head movements, is put forward to alleviate the restriction on the user’s head motion. We manage to improve the performance of the algorithm: The thresholds for extraction of the pupil and the corneal reflection are automated; the edge points of the pupil could be obtained by a 1D line search. To improve the resolution and accuracy of the system, the sub-pixel technique is used to locate the feature position accurately in the eye image. We apply a new randomized method of ellipse fitting to the edge points for good results. Our system can work under natural head movements by using only one camera to track eye-gaze. It can run in real time at video rate (25frames/second). A gaze accuracy of 1 degree is observed.

As shear-warp volume rendering is the fastest rendering method, image quality is not good as that of other high-quality rendering methods. In this paper, we propose two methods to improve the image quality of shear-warp volume rendering. First, the supersampling is performed in an intermediate image space. Then is proposed an efficient method to transform between the volume and the image coordinates at the arbitrary ratio. Second, the pre-integrated rendering technique is utilized for shear-warp rendering. To do this, a new data structure called is used. Using this structure, the empty space leaping can be performed so the rendering speed is maintained even though when the pre-integrated rendering is applied. Consequently, shear-warp rendering can generate on high-quality images comparable to those generated by the ray-casting without degrading the speed.

This paper proposes a simple, efficient method for real-time smoke shadow simulation. In this method, 3D density fields of smoke are directly projected and accumulated into a shadow buffer. Diffusion scheme is creatively taken from fluids dynamics to “soft” shadow texture. These two steps are called Direct-Projection-Diffusion (DPD) which is computationally inexpensive in comparison with traditional methods for participating media volume rendering. For shadow interaction, projective texturing based on the programmable graphics hardware is used to project the shadow texture onto any object’s surface in the scene. The test result shows that our method is efficient and is feasible to solve the problem of real-time smoke shadow simulation in 3D computer games and in 3D animation initial production.

Linux is not well suited for real-time environment. Redesigning it for real-time systems, however, is a very complex and challenging task. In this paper, we suggest an approach to convert a large system like Linux into a real-time system with minimal changes. We observe that most of real-time systems demand real-time performance only for a couple of devices they are targeting for. Our approach focuses on this target device and modifies the original system only where this device is involved. This approach is more practical in that it does not unnecessarily change the whole system and in that it still satisfies the demands of most real-time systems. We picked network device as an example. We assumed the device sometimes got a real-time packet which should be processed immediately.

With the development of semiconductor industry, the chemical mechanical polishing technology has already become the mainstream method of realizing the surface global flatness. In order to understanding physical essence underlying this technology, the author carried out nanometer polishing experiment of silicon wafer using molecular dynamics (MD) simulation method. The simulation result shows that using larger slurry grain can generate much more vacancy, dislocation and larger residual stress than using of small one although using larger slurry grain can acquire better surface quality.

Electroencephalograms (EEG) can provide a unique window on the human brain. Since contamination of EEG recording with artifacts (e.g., signals caused by muscular activity, eye movements, cardiac rhythm and power noise etc.) can decrease the efficiency of diagnosis procedure, here we apply a kind of fast independent component analysis (ICA) approach to analyze the real multi-variant EEG recorded signals. Besides, the comparison between ICA and a second order statistical algorithm are given in this paper. By the real measured data, our experimental results confirm the validity and usefulness of ICA algorithm.

In this paper, MM-Model is presented for real-time simulation of 3D deformable objects on both global level and local region. This model consists of a deformable centerline and dynamic surface reconstruction mechanism based on Mass-Spring and Medial-Representation respectively. When a relatively small force is applied on the object the model works in the same way as the traditional Mass-Spring. Otherwise the force is directly transferred to the centerline and the surface is dynamically recreated according to the position of the centerline. This model works more effectively and efficiently than traditional elastic ones on the global level due to the advantages of the Medial-Representation reflecting the internal information and the Mass-Spring reducing the response time. A novel collision detection algorithm based on adaptive spatial hash, a cutting approach and suture method are also articulated. An artificial blood vessel’s deformation effect and surgery processes are presented in our case study.

The dynamic panorama keeps the advantage of providing both full view of scene in static panoramas and the dynamics of the scene, which remarkably strengthens the reality of walkthrough. This paper presents a method to create dynamic panoramas. To gain the static panorama, a multi-resolution mosaic algorithm based on Particle Swarm Optimization (PSO) is first applied to a series of photographs at one fixed point. Moving objects in the scene are then captured periodically or stochastically using a video camera, and the resulted video clips are converted into video texture. Video texture is finally registered with static panorama and combined into a compact representation. A panorama browser is also expanded to play video textures in addition to its original functions.