In through-the-wall radar imaging, targets behind the wall reflect weak electromagnetic signals that are obscured by the strong returns from exterior wall, rendering the detection and classification of indoor stationary targets very difficult. In this paper, a tensor-based subspace method is proposed for wall clutter mitigation. The radar signals received from the antenna array are transformed into a data tensor. Higher-order singular value decomposition is used to segregate the wall reflections from the target returns. Simulation and experimental results show that the proposed method is effective in removing reflections backscattered from both homogeneous and heterogeneous walls.