FIB-SEM图像多策略层间配准与插值算法  

SLICE REGISTRATION AND INTERPOLATION ALGORITHMS ON FIB-SEM IMAGE WITH MULTI-STRATEGIES

作　　者：张耀 滕奇志[1] 卿粼波[1] 吴晓红[1] Zhang Yao;Teng Qizhi;Qing Linbo;Wu Xiaohong(Institute of Image Information, College of Electronics and Information Engineering ,Sichuan University, Chengdu 610065 ,Sichuan, China)

机构地区：[1]四川大学电子信息学院图像信息研究所,四川成都610065

出　　处：《计算机应用与软件》2018年第5期229-235,共7页Computer Applications and Software

基　　金：国家自然科学基金项目(61372174)

摘　　要：聚焦离子束扫描电镜(FIB-SEM)生成的切片图层间距过大,导致重建后的三维微纳米孔隙图多存在断层现象。为解决此问题,提出一种基于非下采样轮廓波变换NSCT(Nonsubsampled Contourlet Transform)的多策略层间配准与插值算法。该方法利用NSCT捕捉切片图轮廓信息,生成相应NSCT系数;对NSCT低频系数利用基于自由形变FFD(Free-Form Deformation)配准的插值算法生成新插值图的低频子图概貌;对NSCT高频方向子带采用多项式拟合方法进行插值以获得精确的新插值图轮廓结构信息;通过逆NSCT,生成新层间图像,缩减切片纵向间距,提高三维微纳米孔隙图的质量。实验结果表明,该算法可有效提升新层间图像的质量与轮廓结构准确性,削弱噪声影响。相较于其他算法,该算法生成的新层间图像质量在主观与客观上均有所提升。The layer spacing of slices,constructed by focused ion beam scanning electron microscope（ FIB-SEM）,is too large,which leads to fault phenomenon in reconstruction of three-dimensional（3 D） micro-nano pore. To solve the problem,a slice registration and interpolation algorithm with multi-strategies is proposed based on non-subsampled contourlet transform（ NSCT）. NSCT was used to capture contour information and generate the NSCT coefficients; for the low frequency coefficients,interpolation method with FFD-based registration was utilized to generate the low frequency sub-image of new slice; for the high frequency directional sub-bands,polynomials fitting method was made use of generating the accurate contour structure information of new slice; the new slice image was generated by inverse NSCT,thus reducing the layer spacing of the slice and improving the quality of image of the 3 D micro-nano pore. The experimental results indicate that the proposed algorithm can efficiently improve the quality and accuracy of new slice images,and eliminate noises to some extent. Compared with other algorithms,the proposed algorithm can generate more accurate new slice images.

关 键 词：NSCT 图像配准 多项式拟合 FIB-SEM 层间插值 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]