This paper presents a nonlinear method---total variation denoising (TVD) method, for impulse signals denoising.The basic idea of TVD is to solve a total variation function optimization problem.Experimental results suggest that the mean squared error (MSE) can not distinguish the results with some falsely identified impulses.Thus, a dual assessment cfitefion incorporating both MSE and false identification power (fid) is proposed.Numerical experiments have shown that the proposed approach outperforms the traditional wavelet denoising (WD) by using the dual assessment criterion.