视差估计下的VR全景图像局部细节特征提取  

Local Detail Feature Extraction of VR Panoramic Image Based on Parallax Estimation

作　　者：张丽敏[1] ZHANG Limin(Anhui Technical College of Industry and Economy,Hefei 230001,China)

机构地区：[1]安徽工业经济职业技术学院,安徽合肥230001

出　　处：《通化师范学院学报》2024年第2期91-97,共7页Journal of Tonghua Normal University

基　　金：安徽省重点教学研究项目(2021jyxm0196)。

摘　　要：受环境背景、视觉偏差等因素的影响,VR全景图像细节特征提取难度较高.为了优化VR全景图像局部细节特征提取效果,提出视差估计下的VR全景图像局部细节特征提取方法 .在均衡VR全景图像灰度值的基础上,去除VR全景图像的噪声.构建人眼视觉模型,检索去噪后的VR全景图像,获取VR全景图像视差信息.通过匹配代价计算、代价聚合计算、视差计算与后处理计算估计视差数值,利用SIFI算法检测并提取VR全景图像局部细节特征.实验数据显示:应用提出方法获得的局部细节特征点检测结果与实际结果保持一致,视差估计误差最小值为0.3 Pixel,证实提出方法具备较好的局部细节特征提取性能.Due to the influence of environmental background,visual bias and other factors,it is difficult to extract detailed features of VR panoramic images.In order to optimize the local detail feature extraction effect of VR panoramic image,a local detail feature extraction method of VR panoramic image based on parallax estimation was proposed.On the basis of equalizing the gray value of VR panoramic image,the noise of VR panoramic image is removed to construct human visual model,retrieve the denoised VR pan-oramic image,and obtain the parallax information of VR panoramic image.The parallax value is estimated by matching cost calculation,cost aggregation calculation,parallax calculation and post-processing calcu-lation,and the local details of VR panoramic image are detected and extracted by SIFI algorithm.The experimental data show that the local detail feature point detection results obtained by the proposed method are consistent with the actual results,and the minimum parallax estimation error is 0.3 Pixel,confirming that the proposed method has good local detail feature extraction performance.

关 键 词：特征提取 VR全景图像 视差估计 局部细节特征 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]