拍摄全景建筑图像自动拼接匹配仿真  

Simulation of Automatic Stitching Matching of Panoramic Building Images

作　　者：雷桐[1] 燕耀[1] LEI Tong;YAN Yao(Art ＆ Engineering College, Xi＇an Polytechnic University, Xi＇an Shanxi 710048, China)

机构地区：[1]西安工程大学艺术工程学院,陕西西安710048

出　　处：《计算机仿真》2018年第4期377-381,共5页Computer Simulation

基　　金：陕西省教育厅2017年专项科学研究计划项目资助(17JK0308)

摘　　要：对全景建筑不同角度图像的自动拼接匹配,能够有效提高对建筑图像的复原。对全景图像的自动拼接匹配,需要提取全景图像特征区域,对匹配块重叠区域链接约束进行局部优化,完成全景图像的拼接匹配。传统方法获得待拼接图像的显著图,将待拼接图像的前景和背景区域分离,但忽略了优化局部区域链接约束,导致拼接效果不理想。提出基于局部边缘密度LED算法的全景图像自动拼接匹配方法。结合视觉传达技术进行全景图像特征区域提取,采用待拼接图像的局部熵差进行特征点匹配搜索,对特征匹配块重叠区域引入形状扭曲约束和块链接约束进行局部优化,获得修正后的待拼接图像,对待配准图像进行图像融合获得全景拼接图像。实验结果表明,所提方法能够有效突出图像边缘细节,实现了尺度变化条件下的全景图像无缝拼接。Traditional method ignores the optimization of link constraint in local area, resulting in the unideal splicing effect. An automatic stitching matching method of panoramic image of LED algorithm based on local edge density is proposed. Combined with visual communication technology, feature region of panoramic image is extracted. The local entropy difference of image to be spliced is used for matching search of feature point. The shape warping constraint and block chaining constraint are introduced into overlap region of feature matching block for local optimi- zation, so as to obtain modified images which will be spliced. Finally, image composite is used for registering images to obtain panoramic splicing image. Simulation results show that the proposed method can effectively emphasize edge details of images, which achieves seamless stitching of panoramic image at scale variation.

关 键 词：全景图像 自动拼接匹配 局部边缘密度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]