基于层级化匹配的视图变形算法  

View morphing algorithm based on hierarchical matching

作　　者：苏育挺[1] 马玉龙[1] 刘安安[1] 高赞[1] 

机构地区：[1]天津大学电子信息工程学院,天津300072

出　　处：《电子测量技术》2013年第11期52-56,共5页Electronic Measurement Technology

基　　金：国家自然科学基金(61100124;61202168;61170239);天津市应用基础与前沿技术研究计划(10JCYBJC25500);2010和2011天津大学自主创新基金项目

摘　　要：为了提高基础矩阵的准确性和改善视图变形的视觉效果,提出一种基于层级化匹配的视图变形算法。首先对左右视图进行SIFT特征提取并使用传统的基于欧式距离的匹配算法进行粗匹配,然后使用基于局部结构的图匹配算法对粗匹配得到的匹配点对进行细匹配,最后使用细匹配得到的匹配结果计算基础矩阵,并通过视图变形算法合成虚拟视图。实验结果表明,少量的误匹配点对对视图变形的视觉效果有较大的影响,同时所提出的层级化匹配方法能够提高匹配的准确率,使计算得到的基础矩阵更加准确、虚拟视图的视觉效果获得了改善。In order to improve the accuracy of fundamental matrix and the visual effects of view morphing,this paper presents a new view morphing algorithm based on hierarchical matching.The algorithm firstly matches the SIFT feature extracted from the left and right view by traditional Euclidean distance-based matching algorithm that is used as the coarse matching.Then,the points obtained above are matched finely by factorized graph matching algorithm.Lastly,it calculates the fundamental matrix with the points obtained by the fine matching,and synthesizes a virtual view through view morphing.The experimental results show that few mismatched points can have a great impact on the visual effects of view morphing and the proposed method,which has improved the accuracy of matching,makes the fundamental matrix calculated more accurately and visual effects of the virtual view better.

关 键 词：基于图像的绘制 视图变形 基础矩阵 SIFT特征 基于局部结构的匹配 层级化匹配 

分 类 号：TN391.41[电子电信—物理电子学]