THz time-domain spectroscopy system (THz-TDS) can provide spectral information about materials, but the data acquisition takes a long time in THz imaging applications. The compressed sensing algorithm is a computational imaging technique. In this paper, we present a THz compressed sensing imaging method based on an improved linearized-ADMM algorithm. Linearized ADMM methods linearize access to the optimal solution through their gradient evaluations and can effectively solve the sparse recovery problem. The proposed algorithm needs much fewer THz spectra and reduces acquisition time. First, the Gaussian filter in the image domain and the Nesterov acceleration method are adopted in the iterative process to facilitate the convergence speed. Secondly, the MAD (Median absolute deviation) is chosen to calculate the threshold in the soft-shrink step. MAD is a robust measure of sample differences in univariate data sets. It is more flexible than the standard deviation in handling outliers in data sets, and can greatly reduce the impact of outliers on data sets. Thirdly, the L1-reweighted algorithm and averaging sparsity technique are used to enhance the performance and stability of the proposed approach. Finally, the THz beam pattern is taken into consideration to improve image resolution. The experimental results demonstrate that the proposed method has a good performance in THz imaging.