Pan Zhigeng

Semantic modeling project: building vernacular house of southeast China

In this paper, we introduce a semantic modeling project by implementing a southeast China vernacular urban modeling system. This system converts the basic modeling components of geometry units (e.g. point, line, triangle etc.) into the semantic components (e.g. street, block, patch, house etc.). To implement the urban modeling system, an improved production based grammar system is used to control the process of urban generation in detail, and this improved grammar system can avoid the disadvantage (such as hard to control etc.) of the grammar system and to ensure the coherent architecture styles of ancient vernacular houses. Because of the advantages (accurate modeling and rapid generating) of this semantic project system, it can be applied to the preserving of the digital heritage in southeast China vernacular architecture.In this paper, we introduce a semantic modeling project by implementing a southeast China vernacular urban modeling system. This system converts the basic modeling components of geometry units (e.g. point, line, triangle etc.) into the semantic components (e.g. street, block, patch, house etc.). To implement the urban modeling system, an improved production based grammar system is used to control the process of urban generation in detail, and this improved grammar system can avoid the disadvantage (such as hard to control etc.) of the grammar system and to ensure the coherent architecture styles of ancient vernacular houses. Because of the advantages (accurate modeling and rapid generating) of this semantic project system, it can be applied to the preserving of the digital heritage in southeast China vernacular architecture.

A new mesh simplification algorithm based on triangle collapses

In this paper a new mesh simplification algorithm based on triangle collapses is presented. The algorithm can provide efficient error management and simplify the original mesh greatly. Progressive meshes may be constructed with triangle collapsing operation. To make continuous transition between level of detail (LOD) models possible, a method for interpolating is also presented. Examples illustrate the efficiency of the algorithm.In this paper a new mesh simplification algorithm based on triangle collapses is presented. The algorithm can provide efficient error management and simplify the original mesh greatly. Progressive meshes may be constructed with triangle collapsing operation. To make continuous transition between level of detail (LOD) models possible, a method for interpolating is also presented. Examples illustrate the efficiency of the algorithm.

The 3D caricature face modeling based on aesthetic formulae

Recently automatic 3D caricature face generation has attracted much attention from both the research community and the game industry. With the developing of computer vision and modeling, it has become the key technique of virtual avatar and computer games. Let the cartoon face resemble the realistic picture to be customized avatar standing for the user himself. Then the cartoon face also has some exaggeration that can protect the privacy while transmitting through internet. Although it has received considerable attention in computer graphics, due to the lack of 3D caricature sample that many current 3D caricature generation systems required, it still remains as a challenge to generate a satisfactory caricature face model. 3D cartoon face model can be generated from an ordinary frontal face photograph in this paper. We first use an improved ASM to detect the face feature points and receive the different face components region and their distribution. Then transform the face components shape and face contour according to the mean art exaggerate formulae by manifold RBF. The face texture can be transformed from real picture to cartoon one by PCA. Finally, we implement a prototype system that can transform the ordinary face picture into some real and caricature model that can be applied successfully in multi-user educational game to provide human-like avatars. The effectiveness of the proposed technique is demonstrated by the experimental results.

Watermarking T-Spline surfaces

This paper proposes two data embedding algorithms for T-spline surfaces, one is suitable for steganography, and the other for robust shape-preserving watermarking. For steganography, we embed data into the points on T-spline surfaces by using constrained optimization method. For robust shape-preserving watermarking, we use the local knot insertion of T-spline surfaces. The experiment results and their analyses are also presented.

New texture morphing method for visual presentation of textile product

Textile emulation is a key problem in the visual textile presentation. In this paper, we present a new algorithm to achieve the realistic feeling of textile. This algorithm first divides the target area into some texture blocks that have different texture directions. Then morphs the texture coordinate of each texture block. The calculating speed of this algorithm is very fast, so it is more suitable for the visual textile presentation of e-commence.

Multi-channel train visual simulation system based on PC cluster

Train visual simulation system is an important component of a train simulator. Now advanced train visual simulation systems are usually equipped with large-wide projection screen or multiprojection screens to display, and high-end graphic workstation to complete the multiple channels real-time rendering task. They can provide the user large field of view and bring a strong feeling of immersion with the support of interactive devices. This paper presented a multi-channel train visual simulation system based on PC cluster. The new system has a close or even better performance compared to SGI high performance workstation based visual simulation system, furthermore it will cut down the system cost a lot. The techniques of photorealistic rendering, emergency simulation and natural phenomena simulation such as raining and snowing are also discussed in detail.Train visual simulation system is an important component of a train simulator. Now advanced train visual simulation systems are usually equipped with large-wide projection screen or multiprojection screens to display, and high-end graphic workstation to complete the multiple channels real-time rendering task. They can provide the user large field of view and bring a strong feeling of immersion with the support of interactive devices. This paper presented a multi-channel train visual simulation system based on PC cluster. The new system has a close or even better performance compared to SGI high performance workstation based visual simulation system, furthermore it will cut down the system cost a lot. The techniques of photorealistic rendering, emergency simulation and natural phenomena simulation such as raining and snowing are also discussed in detail.

High Quality Range Data Acquisition Using Time-of-Flight Camera and Stereo Vision

Time-of-flight range camera (referred to as TOF camera) has many advantages: compact, easy to use and can obtain three-dimensional depth data of any scene in real-time, which makes it increasingly widely used in various applications. However, TOF camera produces very noisy depth map and often performs poorly on the rich-textured scenes, such as textiles, in which situations stereo vision excels. To solve this problem and use the merits of two methods respectively, we propose a method for jointly using TOF cameras and stereo vision to produce a high quality depth map for surfaces that have rich texture. We choose textiles as experimental object, and the result shows that our method significantly improves the quality of the textiles depth data captured from TOF camera. So we also expand the scope of application of TOF cameras.

Digital Screening Algorithm Based on Improved Delaunay Triangulation

This paper proposes an improved Delaunay triangulation algorithm for digital screening. The algorithm can give a real-time result for its high efficiency. It can give a good solution for Moire patterns phenomenon when printing in traditional pattern. Output image is accurate and soft since gray value distributes uniformly. It has a low request for equipment and subsequent handling, and entirely meet the needs of traditional printing technology.

Spherical harmonic descriptor for gray-level image similarity matching

In this paper, we present a new spherical harmonics-based descriptor for matching the shape similarity between two gray-level images. We describe the properties of this descriptor and show how it can be applied to match the shape similarity between two gray-level images. The advantages of this descriptor are two-fold. First, the spherical harmonic descriptor is invariant to rotation. Second, the spherical harmonic descriptor is a 3D descriptor, which in turn makes comparison of the two gray-level images more efficient.In this paper, we present a new spherical harmonics-based descriptor for matching the shape similarity between two gray-level images. We describe the properties of this descriptor and show how it can be applied to match the shape similarity between two gray-level images. The advantages of this descriptor are two-fold. First, the spherical harmonic descriptor is invariant to rotation. Second, the spherical harmonic descriptor is a 3D descriptor, which in turn makes comparison of the two gray-level images more efficient.

3D Model Retrieval Based on Fuzzy Correspondences and Hybrid Shape Features

With the emergence of a large number of 3D models, retrieving 3D models is becoming more and more important. It is well known that 3D shape features suffer from sparse sampling and view-based features have the ambiguity of different views for a 3D model. In this paper we present a method which combines 3D shape features with view-based features of 3D model to retrieve models. Moreover, according to the retrieval results by using global features search, we provide a retrieval refinement based on local features correspondences. The results show our method is superior to others in different data sets.