SURE-DSS, a method for blind separation of radar signals with noise was proposed based on Stein's Unbiased Risk Estimate Singular Value Thresholding (SURESVT) and Denoising Source Separation (DSS). First, SURESVT was used in place of the singular value decomposition of DSS for obtaining the optimum threshold of the singular value for the observed data, under Stein's unbiased risk estimation principle. Then, the singular values of the observed data were compressed to improve the signal-to-noise ratio while implementing data whitening. At last, the whitened data was separated blindly. The simulation results show that the proposed method can separate the mixed signals effectively for the array model with noise.