目的基于网格变形的图像配准方式,针对待拼接图片重叠区域的视差具有一定的容忍性,并且能够适应更复杂的图像拼接场景。在NISwGSP(natural image stitching with the global similarity prior)算法基础上提出了一种具有直线结构保护的...目的基于网格变形的图像配准方式,针对待拼接图片重叠区域的视差具有一定的容忍性,并且能够适应更复杂的图像拼接场景。在NISwGSP(natural image stitching with the global similarity prior)算法基础上提出了一种具有直线结构保护的图像拼接算法(MISwLP),该算法通过提取图片中的直线结构并施加约束,可以得到视觉效果自然、畸变较小的图像拼接结果。方法首先对图片进行网格划分,建立网格优化模型,针对网格顶点坐标集定义能量函数,在保证图片重叠区域高度对齐的同时,对网格进行相似性连续约束,并辅以直线结构约束,最后使用共轭梯度最小二乘法求解得到最优网格顶点集,指导网格变形。结果针对不同场景下的图片进行拼接实验,同时和几种比较流行的图像拼接软件和算法进行比较。结果表明,同经典拼接算法,比如Autostitch相比,基于网格优化的图像拼接算法能够适应更加复杂的多平面场景,在减小投影失真和对齐误差方面表现更好;同现在比较好的几种网格拼接算法,比如SPHP(shape-preserving half-projective warps for image stitching)、APAP(as-projective-as-possible image stitching with moving DLT)、NISwGSP等的比较,MISwLP算法不仅能够很好地对齐图像和避免投影失真,并且能够保持图像重叠区域到非重叠区域的一致性,即保护原图中的直线结构。结论提出了一种基于网格优化的直线约束方法,对于具有显著几何结构的图像拼接场景,能够较好地保护拼接后图像中原有的直线结构,具有较好的应用价值。展开更多
Objectives： This article presents a new computational procedure to discover scratches buried in the earth＇s crust. We also validate this new interdisciplinary analysis method with regional gravity data located in a ...Objectives： This article presents a new computational procedure to discover scratches buried in the earth＇s crust. We also validate this new interdisciplinary analysis method with regional gravity data located in a well-known Dabie orogenic zone for test. Methods： Based on the scratch analysis method evolved with mathematical morphology of surfaces, we present a procedure that extracts information of the crustal scratches from regional gravity data. Because the crustal scratches are positively and highly correlated to crustal deformation bands, it can be used for delineation of the crustal deformation belts. The scratches can be quantitatively characterized by calculation of the ridge coefficient function, whose high value traces delineate the deformation bands hidden in the regional gravity field. In addition, because the degree of crustal deformation is an important indicator of tectonic unit divisions, so the crust can be further divided according to the degree of crustal deformation into some tectonic units by using the ridge coefficient data, providing an objective base map for earth scientists to build tectonic models with quantitative evidence. Results： After the ridge coefficients are calculated, we can further enhance the boundary of high ridge-coefficient blocks, resulting in the so-called ridge-edge coefficient function. The high-value ridge-edge coefficients are well correlated with the edge faults of tectonic units underlay, providing accurate positioning of the base map for compilation of regional tectonic maps. In order to validate this new interdisciplinary analysis method, we select the Dabie orogenic zone as a pilot area for test, where rock outcrops are well exposed on the surface and detailed geological and geophysical surveys have been carried out. Tests show that the deformation bands and the tectonic units, which are conformed by tectonic scientists based on surface observations, are clearly displayed on the ridge and ridge-edge coefficient images obtained in this article. Moreove展开更多
文摘目的基于网格变形的图像配准方式,针对待拼接图片重叠区域的视差具有一定的容忍性,并且能够适应更复杂的图像拼接场景。在NISwGSP(natural image stitching with the global similarity prior)算法基础上提出了一种具有直线结构保护的图像拼接算法(MISwLP),该算法通过提取图片中的直线结构并施加约束,可以得到视觉效果自然、畸变较小的图像拼接结果。方法首先对图片进行网格划分,建立网格优化模型,针对网格顶点坐标集定义能量函数,在保证图片重叠区域高度对齐的同时,对网格进行相似性连续约束,并辅以直线结构约束,最后使用共轭梯度最小二乘法求解得到最优网格顶点集,指导网格变形。结果针对不同场景下的图片进行拼接实验,同时和几种比较流行的图像拼接软件和算法进行比较。结果表明,同经典拼接算法,比如Autostitch相比,基于网格优化的图像拼接算法能够适应更加复杂的多平面场景,在减小投影失真和对齐误差方面表现更好;同现在比较好的几种网格拼接算法,比如SPHP(shape-preserving half-projective warps for image stitching)、APAP(as-projective-as-possible image stitching with moving DLT)、NISwGSP等的比较,MISwLP算法不仅能够很好地对齐图像和避免投影失真,并且能够保持图像重叠区域到非重叠区域的一致性,即保护原图中的直线结构。结论提出了一种基于网格优化的直线约束方法,对于具有显著几何结构的图像拼接场景,能够较好地保护拼接后图像中原有的直线结构,具有较好的应用价值。
基金National Science Foundation and Chinese Geological Survey for supporting this work
文摘Objectives： This article presents a new computational procedure to discover scratches buried in the earth＇s crust. We also validate this new interdisciplinary analysis method with regional gravity data located in a well-known Dabie orogenic zone for test. Methods： Based on the scratch analysis method evolved with mathematical morphology of surfaces, we present a procedure that extracts information of the crustal scratches from regional gravity data. Because the crustal scratches are positively and highly correlated to crustal deformation bands, it can be used for delineation of the crustal deformation belts. The scratches can be quantitatively characterized by calculation of the ridge coefficient function, whose high value traces delineate the deformation bands hidden in the regional gravity field. In addition, because the degree of crustal deformation is an important indicator of tectonic unit divisions, so the crust can be further divided according to the degree of crustal deformation into some tectonic units by using the ridge coefficient data, providing an objective base map for earth scientists to build tectonic models with quantitative evidence. Results： After the ridge coefficients are calculated, we can further enhance the boundary of high ridge-coefficient blocks, resulting in the so-called ridge-edge coefficient function. The high-value ridge-edge coefficients are well correlated with the edge faults of tectonic units underlay, providing accurate positioning of the base map for compilation of regional tectonic maps. In order to validate this new interdisciplinary analysis method, we select the Dabie orogenic zone as a pilot area for test, where rock outcrops are well exposed on the surface and detailed geological and geophysical surveys have been carried out. Tests show that the deformation bands and the tectonic units, which are conformed by tectonic scientists based on surface observations, are clearly displayed on the ridge and ridge-edge coefficient images obtained in this article. Moreove
