A novel denoising method for mechanical vibration signals is proposed based on quantum superposition inspired parameters estimation. Considering relation between dual-tree complex wavelet transform real coefficients and imaginary coefficients, a new two-dimension probability density function with an adaptive parameter is built up. Investigating the inter-scale dependency of coefficients and their parents, the quantum superposition inspired probability of signal and noise is presented. Integrating the Bayesian estimation theory, an adaptive shrinkage function is infered based on quantum superposition inspired parameters estimation. At last, simulated signals and rolling bearing fault vibration signal are analyzed, which shows that our method can reduce noise effectively, can achieve much better performance than that of the soft and hard threshold denoising methods.