期刊文献+

顾及立面元素排列特性的建筑物纹理修复方法

Building Texture Repair Algorithm Considering Facade Elements Arrangement
下载PDF
导出
摘要 纹理映射是实景三维重建中重要的技术组成部分,可以显著地增加建模场景的细节和真实感。在自动纹理映射后,由于模型之间的相互遮挡导致的错误纹理贴图需要进行人为修复。而现有的图像修复方法会出现修复区域模糊和与周围区域结构衔接错位等问题。为了解决这些问题,提出了一种顾及立面元素排列特性的立面纹理修复方法,该方法分为自适应层高计算和最佳偏移量搜索两部分,立面层高由自动检测出的引导线对计算得到,而最佳偏移量基于立面层高与修复区域周围特征得到,二者结合可以找到立面图像上错误映射区域最佳的纹理替换位置。最后,利用泊松算法对修复边缘纹理进行融合处理,以提高修复区域与周围区域的色调匹配度。实验表明,该方法可以有效利用立面元素的排列特性修复纹理破损区域,与现有图像修复方法比较,该方法在自动化程度及修复效果上均获得了更好的结果。 Texture mapping is an important technical component in 3D reconstruction of real scenes,which can significantly increase the detail and realism of the modeled scene.After automatic texture mapping,incorrect texture maps due to mutual occlusion between models need to be manually fixed.However,the existing image inpainting algorithms will result in blurred object area and incoherent connection with the surrounding known area structure.To solve these problems,this paper proposes a facade texture repair algorithm according to the regular arrangement of facade elements.The algorithm is divided into two parts:adaptive layer height calculation and optimal offset search.The feature line pair is calculated,and the optimal offset is obtained based on the elevation layer height and the surrounding features of the repaired area.The combination of the two can find the best texture replacement position for the incorrectly mapped area on the elevation image.Finally,the inpainted edge textures are fused by Poisson algorithm to improve the tonal matching between the inpainted area and the surrounding area.Experiments show that the algorithm can effectively use the arrangement characteristics of facade elements to repair damaged texture areas.Compared with the existing image repair methods,this method achieves better results in terms of automation and repair effect.
作者 龚云璐 李大军 郭丙轩 刘波 GONG Yunlu;LI Dajun;GUO Bingxuan;LIU Bo(Faculty of Geomatics,East China University of Technology,330013,Nanchang,PRC;State Key Laboratory of Information Engineering in Surveying,Mapping and Remote Sensing,Wuhan University,430079,Wuhan,PRC)
出处 《江西科学》 2022年第5期997-1003,共7页 Jiangxi Science
基金 国家自然科学基金项目(42161064) 江西省自然科学基金项目(20212BAB204003)。
关键词 图像修复 立面纹理修复 自适应层高计算 最佳偏移量搜索 立面元素排列特性 image inpainting facade texture repair adaptive layer height calculation best offset search arrangement properties of facade elements
  • 相关文献

参考文献2

二级参考文献13

共引文献15

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部