【作者】 赵国英；

【导师】 黄心渊；

【作者基本信息】 北方工业大学 ， 计算机应用， 2002， 硕士

【摘要】 我们的课题涉及的是VR领域的立体化研究，它是VR(Virtual Reality)研究领域的一个组成部分。虚拟现实的视觉显示系统大致可以分为3大类：头盔式显示屏、高分辨率CRT系统和与传统显示器连用的三维立体眼镜。就目前我国的情况而言，后两种系统可能逐渐在民用领域得到广泛应用。但是无论哪种系统，都必须解决立体显示和立体图像资源问题。立体显示技术(也称为体视技术、视差技术)是相对传统显示技术而言，它主要借鉴人类的视觉原理，使平面图像能产生出真立体效果。现在软件及作品问题已经是阻碍实用VR普及的关键问题之一。我们知道，如果没有人们感兴趣的作品，再好的观察设备也是难以推广的。就如同没有软件，计算机也会没有市场一样。 立体化研究工作大致可以分为两个方面：一是硬件的研制与开发，如观察设备就包括计算机上VR观察设备，电视上的观察设备和简单的图片观察设备等。二是软件的研制与开发，包括体视图像的生成算法以及高效快速生成体视视频的技巧与方法等。本文的工作内容主要集中在第二方面，包括：1 完善了体视生成算法 在平移算法的基础上，提出了柱面投影和球面投影算法，把二维平面图像或视频从两个投影点投影到柱面或球面上，以得到体视图像或视频，并在理论上和实际中对各种算法进行了比较。2 完成“体视压缩软件”的设计与开发 根据体视生成算法，利用面向对象程序设计(Object-Oriented Programming简称OOP)，完成了体视压缩软件，可以对BMP、AVI文件进行体视处理，生成体视BMP、AVI文件或经过MPEG2编码后的体视DVD文件。3 利用体视生成软件制作了一些好的真立体节目 对3DS MAX生成的动画“申奥”和影片“MATRIX”等进行了体视处理。4 分层体视初步实现 我们还在分层体视生成方面进行了初步实验和探索，把图像分成不同层，对不同层 北方工业大学学位论文运用不同体视算法或设置不同参数来进行体视转换，来得到合成体视图像。 我们在图像分割和裸眼可视方面进行了理论上的探讨，最后对本文所做工作进行总结并讨论将来可能继续的研究方向。更多还原

【Abstract】 The research we are involved in is strereoscopiclization in VR technology, which is a component of VR research areas. The vision display system of VR can be approximately divided into three categories: Head-Mounted Displays (HMDs), high resolution CRT system and three-dimensional stereoglasses connected with traditional display. As far as it goes, the latter two systems will be widely applied in civil area. But no matter which system is used , we must settle the problem of stereoscopic display and stereoscopic image source. Compared with traditional display technology, the stereoscopic display technology (namely stereoscopic technology, parallax technology) mostly uses human vision principle for reference, and makes the flat image have the true solid impression, that is to say, makes the flat image have depth of field sense according to the parallax of two eyes. Now software and production are one of the key problems blocking the spread of practice VR. As we all know, if there is not the production that people are interested in, viewing device cannot be expanded no matter how excellent it is, just as computers can not find a good market without softwares.There are approximately two aspects of stereovision study: one is the research and development of hardware, such as the viewing equipment including the VR viewing device on computers, on TVs and the simple image viewing device. The other is the study and development of software, which includes the generation algorithm of stereoscopic image, the skill and method of fleetly generating stereoscopic videos with high efficiency. In this thesis, our research work mainly focuses on the second aspect, including: 1 Perfecting the generation algorithms of stereoscopic image and videoOn the basis of moving flatly, we brought forward the cylindrical projection and spherical projection algorithms, projecting the flat image or video from two viewing points to the cylinder or sphere to get the stereoscopic image or video, and compared these three algorithms based on the theory and practice.I2 Completing the design and development of the "Generation and Compression System of Stereoscopic Video"According to the stereoscopic generating algorithms, we completed the "Generation and Compression System of Stereoscopic Video" with Object-Oriented Programming method. This system can process the BMP and AVI files to generate the stereoscopic BMP and AVI files or the stereoscopic DVD files which are encoded with MPEG2(Moving Picture Experts Group) standard.3 Making some excellent true stereoscopic programs with our "Generation and Compression System of Stereoscopic Video"We used the stereoscopic algorithms to process the "Olympic Green" animation which was generated by 3DS MAX and the famous film "MATRIX".4 Elementary implementation of layered stereoscopic image and effect comparisonWe explored the layered stereoscopic generation on the basis of object segmentation. Before the stereoscopic transform, we process the image to make it layered, then apply different methods and set different stereoscopic parameters to different layers. Finally, we synthesize these processed layers to get a whole stereoscopic image. We compared original stereoscopic image with the synthesized stereoscopic images generated on the basis of layered images.We theoretically discuss the object segmentation and autostereoscopic technology. At last, I summarize the work of this thesis and discuss the potential research topics in the future.更多还原