近年来，数字全息显微技术在复杂表面测量领域得到了广泛关注。数字全息技术主要分为干涉记录和重构再现两部分。在激光干涉记录中，散斑噪声会严重影响测量结果。因此数字全息相位重构的一项重要工作就是去除散斑噪声，以提高重构精度。分析了带有散斑噪声的几类典型的结构面，比较了菲涅尔衍射算法、基于小波变换的菲涅尔衍射算法和新提出的基于非下采样轮廓波变换的菲涅尔衍射算法三者在散斑去噪和信号保持方面的表现。结果显示：基于非下采样轮廓波变换的菲涅尔衍射重构算法可以有效提取各类结构面的信息，有效实现信噪分离。因此，该方法比现有重构算法可得到更高的重构精度，在结构表面的相位重构方面具有重要应用价值。“,”In recent years, digital holographic microscopy has attracted intensive attention for its capability of measuring complex shapes. There are two parts in digital holographic microscopy, hologram recording and digital diffractive reconstruction. Speckles are inevitable in the recorded interferometric patterns, thereby polluting the reconstructed surface topographies. Three reconstruction algorithms, i.e. Fresnel transform, Fresnel-wavelets and the proposed Fresnel-NSCT algorithms were compared. Three typical structures, rectangular, spherical and triangular surfaces were adopted for analysis. The performance of the three reconstruction algorithms on speckle removal and feature preservation was investigated comprehensively. Signal-Noise-Ratio (SNR) and Peak-Signal-Noise-Ratio (PSNR) were used as the numerical criteria. It is found that the Fresnel-NSCT algorithm has great superiority over the other two, subsequently it is promising for applications in the diffractive reconstruction of structured surfaces.