太赫兹同轴数字全息再现像噪声较大，尤其是背景噪声,严重影响着图像整体效果。因此,为实现图像平滑,提出了联合数学形态中重建的开运算和闭运算的非局部均值或中值非局部均值滤波进行太赫兹Gabor同轴数字全息二维再现像的图像复原方法，并对目标分辨率较低的真实再现像进行了数字处理比较分析。所研究的图像处理结果表明，提出的复合算法能够去除绝大部分噪声，达到图像平滑和提高信噪比的目的，效果明显优于单独使用复合算法中任一滤波方法。

The reconstructed image noises of THz in-line digital hologram are higher, especially the background noises, which seriously affect the overall effect of the image. To realize image smoothing, this paper proposes the restoration methods on two-dimensional reconstructed images of THz Gabor in-line digital hologram by combining opening-closing-by-reconstruction in mathematical morphology with non-local means or median non-local means filter. Digital processing and comparative analysis for the real reconstructed image of the target with lower resolution are carried out. The images processing results show that the proposed composite algorithm can remove most of the noise, leading to image smoothing and signal-to-noise ratio improvement, and the effect is obviously better than seperately using any one single filter in composite algorithm.