摘要随着二维图像采集、编码、传输和显示的不断发展及完善,物体的二维信息已经变得远远不能够满足人类的需求,人们现在对再现真实物体的需求已经要求要达到再现景物的三维信息。所以国内外相关领域的研究热点已经转向研究立体成像的技术，而立体成像技术的其中两个重要分支则是多视点自由立体图像实时获取技术和多视点自由立体图像显示技术。目前最常采用的立体图像获取技术是用按照规律排列的N个摄像机同步拍摄同一个场景在不同的位置(视点)下的N幅图像,然后利用一些算法对拍摄到的N幅图像进行合成。但是这种方法存在很大的缺点。这种方法要求必须保证N个摄像机同时进行拍摄，并且各台摄像机的控制参数必须完全相同,而且精确度要求得非常高,这一点在多个摄像机拍摄的时候很难实现。因此,多视点立体图像实时获取技术仍然是立体成像技术研究中的难点。64921

本文介绍利用2台或者3台摄像机同步拍摄获得两个视点或者三个视点的图像,然后利用深度图插值或图形变形,生成中间各个视点的虚拟视点的图像,这样就就可以减少同时拍摄的摄像机的数量,只需要保证拍摄的两台或三台摄像机的参数一致就可以,而且对实时性的要求也降低了,因此使得系统的复杂度大大降低,由生成的虚拟视点图像对合成的立体图像,通过主观评价系统的验证,其各项评价的指标都接近实际不用算法生成的直接拍摄合成得到的立体图像,具有很好的显示效果。

毕业论文关键词：立体成像 立体图像实时获取  图形变换  虚拟视点

毕业设计说明书（论文）外文摘要

Title      The research of point  of  view  synthesis     

Abstract

With the continuous development of 2D image acquisition, coding,transmission and display and perfect, two-dimensional information of objects has become far less able to satisfy people's needs, people are now on the reproduction of the real object demand has required to get 3D information reproducing scenery. So the hot spot of research has shifted to the study of stereo imaging and three-dimensional imaging techniques, which are two important branches is the multi-view auto-stereoscopic image real-time acquisition and multi-view auto-stereoscopic display technology. At present the most commonly used stereo image acquisition technology is the same scene in different position synchronous shooting in accordance with the laws of permutations of N camera (point of view) of the N image, then use some algorithm N images on the shooting of synthesis. But this method has a great disadvantage. This approach requires must ensure that N cameras to film, and the control parameters of the camera must be exactly the same, but the accuracy is very high, this point shooting in multiple cameras when it is difficult to achieve. Therefore, multi-view real-time stereo image technology is still difficult to study three-dimensional imaging technology in.

    This paper introduces synchronous images obtained two point or three point of use 2 or 3 cameras, and then use the depth map interpolation or graphic image deformation, virtual view generation each viewpoint, this can reduce the number of camera and shooting, parameters of two or three cameras need only guarantee photography can be the same, but also for the real-time requirement is reduced, so that the complexity of the system is greatly reduced, the virtual viewpoint image generation of stereoscopic image synthesis, validated by subjective evaluation system, the evaluation indicators are close to the actual no algorithm to generate the direct shooting synthetic stereo images obtained, have a good effect.

Keywords: stereo imaging     Real-time stereo image acquisition   Graph transformation    virtual viewpoint