This paper applies the matrix low-rank decomposition technique to the problem of seismic data reconstruction. In order to protect the local structure, total variation regularization was introduced. The problem of missing channel reconstruction is solved by adopting the joint optimization model of matrix low-rank decomposition and total variation regularization, and the model is implemented by using the (Alternating Direction Method of Multipliers, ADMM) algorithm and the mixed optimization algorithm of proximal operators. The experiments with synthetic data and actual data have shown that this model can effectively interpolate the missing parts of seismic data.