Abstarct: Attenuation of random noise is a major concern in seismic data processing. Such noise usually characterized by random oscillation in seismic data over entire time and frequency. We introduced and evaluated a low-rank and sparse decomposition baxsed method for seismic random noise attenuation. The proposed method starts by transforming the seismic trace into a new sparse subspace using a sparse time-frequency representation (TFR). Then the sparse TFR matrix were decomposed in two parts: (a) a low-rank and (b) a sparse components using bilateral random projection (BRP). We can recover the de-noised seismic signal by minimizing mixed 1 2  norms objective function by considering the intrinsically low-rank property of the seismic data and sparsity feature of random noise in the sparse TFR domain. Our method was tested with synthetic and real data. De-noising was carried out by means of two another methods. We demonstrate that our proposed method is an effective, amplitude preserving and robust tool that gives superior results compared to another conventional de-noising algorithm.