Visual Hull Generated Based On Polar Coordinates

With the rapid development of virtual reality and digital entertainment, 3D modeling becomes one of the most important directions in computer graphics. Image Based Modeling (IBM) is an important method of 3D modeling. Compared with other methods such as geometry modeling or acquiring models by 3D-Scanning, although Image Based Modeling can not obtain high precision models, this method can construct photorealistic models efficiently with low cost. So it has the potential to be widely used in the game and entertainment industry.Human is very sensitive to the information of silhouettes. Image Based Visual Hull method is an efficient way to construct 3D models from a set of images by computing shape-from-silhouettes. A visual hull is constructed by using the visible silhouette information from a series of reference images. We present in this paper a new approach to improve efficiency and stability. The experimental results demonstrate that our approach can rapidly generate models with high stability.This thesis consists of five chapters. In the first chapter, we briefly survey the approaches and techniques of Image Based Modeling. In the second chapter, we discuss the camera calibration which is an important step of IBM. To simplify the parameters of a set of cameras to be calculated, we use a check-board and a rotating table. The third chapter gives the detailed description of our reconstruction algorithm. In our approach, the approximated center of object is selected from the references images and a polar-coordinate system is set up for the object. From the silhouettes of references images, we obtain point positions of visual hull. Then, we use the relative position in polar-coordinates to construct the mesh of visual hull. After adaptively re-sampling thevisual hull we get the constructed mesh. In the fourth chapter, we discuss the texture mapping method. Here Polar-coordinates are utilized. Multiple reference images are mapped to 3D models and get promising results. Finally, we draw conclusions and discuss future work in the fifth chapter.