基于三维电视深度图提取算法的研究
发布时间：2018-10-12 14:58
【摘要】：近年来,随着众多优秀3D电影的放映,三维显示技术越来越受到人们的关注,作为受众群体最多的电视技术也开始迈入三维电视技术的新纪元。三维电视相较于传统的二维电视能够更加逼真地显示场景,可以实现多角度自由观看,使观众在观看电视节目时能够产生身临其境的感觉,在娱乐,广告,艺术等众多领域都有很好的应用。三维电视技术是目前研究的热点,国内外很多科研机构都在进行相关的研究工作。 本论文介绍了三维电视的国内外发展以及现状,对三维显示技术的发展以及目前三维电视标准进行了介绍,并对三维电视系统中的节目源制作、编码技术以及三维显示等关键技术进行了论述。鉴于三维电视系统中要传输多个视点的视频信息,为了节省带宽,一般采用“视频信息+深度信息”的编码传输方式,在后端采用DIBR(基于深度图的绘制)技术恢复视频信息。深度图的获取是三维电视系统的关键技术,是国内外的研究热点。深度图可以通过深度摄像机直接获取,也可以通过立体匹配算法获得。由于深度摄像机造价昂贵且精度不高,我们重点研究立体匹配算法。 视差和深度存在一定的几何关系,通过立体匹配算法我们可以得到视差图,在已知摄像机参数下就能计算出深度图。按照生成深度图的不同,立体匹配算法可以分为稀疏深度图提取算法和稠密深度图提取算法。因为三维电视系统后期要采用DIBR技术,只能采用稠密深度图生成算法,包括区域匹配算法以及全局匹配算法。因为三维电视系统对深度图的要求较高,本文提出了一种基于分割的快速而精确的立体匹配算法。通过利用均值漂移算法对图像进行分割,可以很好的对图像进行分割；然后通过区域匹配算法快速获得初始视差；然后进行视差分配,将初始视差分配到各个分割区域；最后进行优化。通过实验,我们可以看出这个算法精度以及实时性都不错。
[Abstract]:In recent years, with the screening of many excellent 3D movies, 3D display technology has been paid more and more attention. As the largest audience, the TV technology has entered a new era of 3D TV technology. Compared with the traditional two-dimensional television, 3D TV can display scenes more vividly, can realize multi-angle free viewing, make viewers feel like they are on the spot while watching TV programs, in entertainment, advertising, Art and many other fields have good applications. Three-dimensional television technology is the focus of current research, many research institutions at home and abroad are carrying out related research work. This paper introduces the development and present situation of 3D TV at home and abroad, introduces the development of 3D display technology and the current standard of 3D TV, and makes the program source in 3D TV system. The key technologies such as coding and three-dimensional display are discussed. In view of the need to transmit video information from multiple viewpoints in a 3D television system, in order to save bandwidth, the coding mode of "depth information of video information" is generally adopted, and the technology of DIBR (drawing based on depth map) is used to restore video information at the back end. The acquisition of depth map is the key technology of three-dimensional television system, and it is the research hotspot at home and abroad. The depth map can be obtained directly by depth camera or by stereo matching algorithm. Due to the high cost and low precision of depth camera, we focus on stereo matching algorithm. There is a certain geometric relationship between parallax and depth. The disparity map can be obtained by stereo matching algorithm, and the depth map can be calculated under known camera parameters. According to the difference of generating depth map, stereo matching algorithm can be divided into sparse depth map extraction algorithm and dense depth map extraction algorithm. Because DIBR technology is adopted in the later stage of 3D TV system, dense depth map generation algorithm, including region matching algorithm and global matching algorithm, can only be used. In this paper, a fast and accurate stereo matching algorithm based on segmentation is proposed because of the high requirement of depth map in 3D TV system. By using the mean shift algorithm to segment the image, the image can be segmented well; then, the initial parallax can be obtained quickly by the region matching algorithm; then, the parallax is allocated to each segmentation region. Finally, the optimization is carried out. Through experiments, we can see that the algorithm accuracy and real-time are good.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP391.41
[Abstract]:In recent years, with the screening of many excellent 3D movies, 3D display technology has been paid more and more attention. As the largest audience, the TV technology has entered a new era of 3D TV technology. Compared with the traditional two-dimensional television, 3D TV can display scenes more vividly, can realize multi-angle free viewing, make viewers feel like they are on the spot while watching TV programs, in entertainment, advertising, Art and many other fields have good applications. Three-dimensional television technology is the focus of current research, many research institutions at home and abroad are carrying out related research work. This paper introduces the development and present situation of 3D TV at home and abroad, introduces the development of 3D display technology and the current standard of 3D TV, and makes the program source in 3D TV system. The key technologies such as coding and three-dimensional display are discussed. In view of the need to transmit video information from multiple viewpoints in a 3D television system, in order to save bandwidth, the coding mode of "depth information of video information" is generally adopted, and the technology of DIBR (drawing based on depth map) is used to restore video information at the back end. The acquisition of depth map is the key technology of three-dimensional television system, and it is the research hotspot at home and abroad. The depth map can be obtained directly by depth camera or by stereo matching algorithm. Due to the high cost and low precision of depth camera, we focus on stereo matching algorithm. There is a certain geometric relationship between parallax and depth. The disparity map can be obtained by stereo matching algorithm, and the depth map can be calculated under known camera parameters. According to the difference of generating depth map, stereo matching algorithm can be divided into sparse depth map extraction algorithm and dense depth map extraction algorithm. Because DIBR technology is adopted in the later stage of 3D TV system, dense depth map generation algorithm, including region matching algorithm and global matching algorithm, can only be used. In this paper, a fast and accurate stereo matching algorithm based on segmentation is proposed because of the high requirement of depth map in 3D TV system. By using the mean shift algorithm to segment the image, the image can be segmented well; then, the initial parallax can be obtained quickly by the region matching algorithm; then, the parallax is allocated to each segmentation region. Finally, the optimization is carried out. Through experiments, we can see that the algorithm accuracy and real-time are good.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP391.41
【参考文献】
相关期刊论文 前2条
1 吕朝辉;董跃;;立体电视研究现状与展望[J];电视技术;2006年08期
2 邓智s，
本文编号：2266567
本文链接：https://www.wllwen.com/wenyilunwen/guanggaoshejilunwen/2266567.html
