针对传统经验模态分解（Empirical mode decomposition，EMD）在边缘易出现分解错误的问题，本文提出一种改进的经验模态分解方法。分别对条纹进行镜像延拓和Gerchberg外插迭代来实现边沿的拓展，有效抑制条纹边沿引起的模态分解错误，提高分解准确度。并将改进的EMD分解方式应用于变形结构光条纹图的分析，有效消除条纹中的背景分布，得到更好的三维面形重建效果。
Aiming at the problem of the edge error caused by the traditional empirical mode decomposition method, an improved empirical mode decomposition method is proposed for eliminating the decomposition error in the edge zones of the signals in this paper, in which, a mirror extension method and Gerchberg extrapolation iteration is introduced to eliminate the decomposition error at edges position, respectively. The improved method can effectively suppress the mode decomposition error caused by the signal edge and improve the decomposition accuracy of the EMD. It is also applied in the analysis of the deformed fringe pattern for eliminating the background components in 3D optical measurement, by which a better reconstructed result of 3D surface can be obtained.
引用本文格式： 李绪琴,陈文静,苏显渝. 基于改进经验模态分解的三维重建[J]. 四川大学学报: 自然科学版, 2018, 55: 0111.复制