Tour-into-picture with Kinect

• 主要作者: Meng, Taing
• 其他作者: Zheng Jianmin
• 格式: Final Year Project
• 語言: English
• 出版: 2014
• 主題: DRNTU::Engineering::Computer science and engineering::Computing methodologies::Computer graphics
• 在線閱讀: http://hdl.handle.net/10356/58953
• 機構: Nanyang Technological University

3D graphic is one of the most popular researches in computer graphic, modeling, rendering, and animation. There are various methods in modeling an object or a scene in 3D, but each method has its pros and cons in term of cost, quality and performance. Among those methods, Tour Into the Picture, developed by Y. Horry et al, uses a spidery mesh interface to make animation from a single image. TIP maps a single image with its vanishing point and projection lines through perspective projection into a front-less box. A virtual camera is used to look into the box by rotating and scaling with mouse controls. TIP is fast in performance and simple to model a 3D scene. However, there is a limitation to the images which are not in perspective projection. The foreground objects are flattened or distorted with the background unless manually masked. In this project, we enhanced the Tour-Into-Picture method by introducing Kinect face-tracking to substitute mouse controls. This could provide users more realistic and immersive interaction by looking at a picture from different angles with hand-free experience. Beyond that, we also introduced Tour-Into-the-Video, explored various mapping methods for non-perspective-projection images, and analyzed some automatic foreground and background segmentation algorithms. In conclusion, this project discussed all the possible combinations in computer graphic and computer vision to build on the existing TIP method. Further research may focus on the automation of foreground objects masking and perspective projection line tracing, which requires the least manual configuration.