摘要
图像深度信息提取是三维图像重建的关键,利用匹配算法可以达到对全景成像可控处理的目的,因此在非周期性提取像素基础上完成全景图像三维重构。首先采用非周期性像素提取解决成像过程中出现的失真问题,构建无失真图像;通过立体视图系统和独立立体视图系统对全景图像重建处理,利用SSD计分准则的块匹配算法,对微透镜记录的两幅图像计算提取,同时运用邻域限制与放松方法深度提取全景图像;并依据最佳视差值及多基线立体匹配算法过滤出更准确信息,以采用差值算法实现更高精度的三维重建。仿真结果表明,所提方法具有较高适用性,计算过程简便、图像特征采集快。
Image depth information extraction is the core of 3D image reconstruction,and the application of matching algorithm can control panoramic imaging.Therefore,it is necessary to complete 3D reconstruction of panoramic image based on non-periodic extraction of pixels.Firstly,the non-periodic pixel extraction was utilized to construct the image without distortion for avoiding the appearance of distortion.Secondly,the panoramic image was reconstructed by stereo view system and independent stereo view system.Then,the two images recorded by micro-lens were extracted by block matching algorithm based on SSD scoring criterion.Meanwhile,according to the neighborhood restriction and relaxation method,the depth of panoramic image was extracted.Then,the more accurate information was obtained after filtering via the combination of the best disparity value and multi-baseline stereo matching algorithm.Eventually,the difference algorithm was used to achieve higher precision 3D reconstruction.The simulation results show that the method has high applicability,simple calculation and fast image feature acquisition process.
作者
汪永吉
WANG Yong-ji(Xiangsihu College of Guangxi University for Nationalities,Nanning Guangxi 530008,China)
出处
《计算机仿真》
北大核心
2021年第2期128-132,共5页
Computer Simulation
基金
广西民族大学相思湖学院2019年度院级科研立项项目(2019YJKY06)。
关键词
非周期性提取
全景图像
三维重构
插值算法
Non-periodic extraction
Panoramic image
3D reconstruction
Interpolation algorithm
