基于虚拟视点绘制的3D图像空洞恢复方法研究  

Research on void restoration of 3D image based on virtual viewpoint rendering

作　　者：吴梦蝶 胡文靓 苗水清 Wu Mengdie;Hu Wenliang;Miao Shuiqing(School of Xi’an Innovation,Yan’an University,Xi’an 710100,China)

机构地区：[1]延安大学西安创新学院数据科学与工程学院,陕西西安710100

出　　处：《无线互联科技》2022年第23期134-136,共3页Wireless Internet Technology

基　　金：陕西省教育厅科研项目,项目编号:21JK0997;延安大学西安创新学院校级科研,项目编号:2020XJKY-1。

摘　　要：随着2D-3D转换技术的发展,3D视频逐步走向家庭,3D观影突破电影院的限制走向移动终端。然而,3D产业的发展依然受限于3D视频源的匮乏。将2D图像作为输入合成3D虚拟视点图像,成为解决3D资源匮乏问题的重要方案。文章对2D-3D技术基于深度图的虚拟视点绘制进行了深入研究,针对虚拟视点绘制技术中3D图像生成过程的空洞问题,借鉴医学中的渗透率的概念,对空洞进行基于参透滤波的信息恢复,从而提高合成的3D图像质量,带来更好的3D视觉体验。With the development of 2D-3D conversion technology, 3D video has gradually moved towards the home,and 3D movie viewing has broken through the restrictions of cinemas to mobile terminals. However, the development of 3D industry is still limited by the lack of 3D video sources. Using 2D images as input to synthesize 3D virtual viewpoint images has become an important solution to solve the problem of lack of 3D resources. The depth map based virtual view rendering of 2D-3D technology is deeply studied. Aiming at the hole problem in the 3D image generation process of virtual view rendering technology, the concept of permeability in medicine is used for reference, and the information recovery based on penetration filter is carried out to improve the quality of synthetic 3D image and bring better 3D visual experience.

关 键 词：虚拟视点绘制 3D图像 空洞 信息恢复 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]