Voxel-based Three-dimensional Metamorphosis and Texture Mapping

碩士 === 國立成功大學 === 資訊工程研究所 === 88 === The rapid advent of computer techniques greatly changes the human life and thought. Many modern technologies are embedded in lots of popular tools or things around us. Especially, computer techniques are the most commonly used and deeply immerged in our daily lif...

Full description

Bibliographic Details
Main Authors: Wen-Yan Chang, 張文彥
Other Authors: Yung-Nien Sun
Format: Others
Language:en_US
Published: 2000
Online Access:http://ndltd.ncl.edu.tw/handle/58697423483881175703
Description
Summary:碩士 === 國立成功大學 === 資訊工程研究所 === 88 === The rapid advent of computer techniques greatly changes the human life and thought. Many modern technologies are embedded in lots of popular tools or things around us. Especially, computer techniques are the most commonly used and deeply immerged in our daily life. Image processing and computer graphics have been applied to many fields, such as animation, videoconference, surgical simulation and so on. The processing results are usually influenced by data type, processing method and object properties. Thus, we usually have to design different techniques to handle different type of data depending on these conditions. In medical image data or data obtained by many other sensors, the acquired data are originally in voxel format. There is rich of important information in this type of data set. However, if we process it with a polygon-based method, the processing will result in some loss of information and become very time consuming since we have to change the format into polygons first. Therefore, we propose two voxel-based techniques for three-dimensional (3D) metamorphosis and texture mapping. These methods are based on our previously proposed flattening algorithm, which is based on the popular chain coding technique. In voxel-based data, we can encode the surface of each horizontal cross section by chain coding. Then we can find the alignment point, cut point and the 3D coordinate of each voxel on the surface from these codes. A 2D flattened image can be obtained quickly by using this flattening algorithm. The flattened image can then be used in the processes of 3D metamorphosis and texture mapping. In 3D metamorphosis, we reduce the 3D data to 2D images and then morph them by image morphing techniques. A sequence of 3D intermediate models can then be obtained after reconstruction. Our method simplifies the feature definition and manipulation from 3D space to 2D image, and thus speeds up the computation. For facial texture mapping, we also flatten the 3D model and find the correspondence between the flattened image and the facial texture image by the warping technique. After these steps, we can map the facial images onto the 3D model. And a realistic facial model is obtained. At last, we combine the 3D metamorphosis with texture mapping to provide a tool for morphing 3D-textured object and show very interesting result. By extending the results of our current study, they can be further applied to surgical simulation, animation, tele-medicine applications, etc.