基于深度图的二维视频三维化的相关技术研究

发布时间：2018-03-28 01:11

本文选题：立体视频　切入点：自适应运动估计　出处：《哈尔滨工业大学》2010年硕士论文

【摘要】： 视频的立体显示是今后信息的重要表现形式之一。立体视频技术的发展将带来视频领域的下一次革命。与平面视频相比,立体视频显示层次分明,给人强烈的视觉冲击,留下深刻的印象;除了更高的视频质量外,立体视频给观众带来身临其境的感觉,也是其越来越受到青睐的原因。正因为立体视频具有平面图像和视频所不具备的特点,所以在诸如电脑显示、电视、测量、三维医学图像重建、虚拟现实、计算机动画、广告传媒、电子游戏等领域具有广阔的应用前景。但是利用特殊设计的摄像机直接产生立体视频的方法,费用是非常高昂的。将平面视频转换为立体视频的技术无疑是一种既经济又便捷的技术。因此,本文利用双目视觉原理,深入地对平面视频立体化的相关技术进行了研究。本文设计了平面视频-立体视频转换系统的结构。系统首先对平面视频序列进行运动估计,再从运动矢量中获取深度信息,这其中引入了互信息量这一概念,提出了适用于深度信息提取的自适应运动估计算法。此算法在降低运动估计运算复杂度的同时,还可以有效提高深度图的质量。接着利用深度信息与双眼视差之间的关系,计算出双眼视差,进而得到左、右眼视图。合成后即获得了具有立体感觉的三维视频图像。总的来说,平面视频-立体视频的转换技术是一个全新的领域,同时也是一个较难的研究课题。通过本文的研究,提出了一套可以实现平面视频立体化的转换系统,并将整个系统在软件平台上加以实现,目前虽然系统的性能尚有许多有待改进之处,但基本已能得到较为令人满意的具有立体感觉的视频效果。软件系统设计实现之后,将来要把整个系统移植到硬件平台上,使其具有实际应用价值,能够在立体视频相关的应用领域方面发挥应有的巨大潜在价值。
[Abstract]:Stereo display of video is one of the important forms of information in the future. The development of stereoscopic video technology will bring the next revolution in the field of video. Impressed; in addition to higher video quality, stereoscopic video gives viewers a sense of reality, which is why it is becoming more and more popular. It is precisely because stereoscopic video has features that plane images and videos do not. So in things like computer display, television, measurement, 3D medical image reconstruction, virtual reality, computer animation, advertising media, Video games and other fields have a broad application prospects, but the use of specially designed cameras to directly generate stereo video, the cost is very high. The technology of converting plane video to stereoscopic video is undoubtedly an economical and convenient technology. In this paper, the structure of plane video-stereo video conversion system is designed. Firstly, the motion estimation of plane video sequence is carried out, and then the depth information is obtained from motion vector. In this paper, the concept of mutual information is introduced, and an adaptive motion estimation algorithm for depth information extraction is proposed, which not only reduces the computational complexity of motion estimation, but also reduces the computational complexity of motion estimation. The relationship between depth information and binocular parallax is used to calculate the binocular parallax, and then the left and right eye views are obtained. After synthesis, three-dimensional video images with stereosensory are obtained. In general, the conversion technology of plane video to stereo video is a new field, and it is also a difficult research topic. Through the research of this paper, a set of conversion system is proposed, which can realize the three-dimensional conversion of plane video. The whole system is implemented on the software platform. Although the performance of the system still needs to be improved, it can get satisfactory video effect with stereosensory. After the software system is designed and implemented, it is necessary to transplant the whole system to the hardware platform in the future, so that it has practical application value and can play a great potential value in the application field of stereoscopic video.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2010
【分类号】：TP391.41

【参考文献】