Abstract: In order to better obtain the original signal under the influence of noise, on the basis of a shrinkage method of neighbouring wavelet coefficient called NeighCoeff, we propose a wavelet de-noising method based on multi-threshold new function with neighborhood correlation(MNFNC). According to the correlative characteristics of wavelet coefficients, MNFNC classifies quadratic sum of all wavelet coefficients in neighborhood window as neighboring hard threshold value, neighboring window threshold value and neighboring expansion threshold value. A comparison between these neighboring threshold values and the revised universal threshold achieves adaptive adjustment of window size and retention or shrinkage of wavelet coefficients, which is designed to remove noise. In addition, we use the shrinkage factors of MNFNC, which well reflect the relationship with the filtered noise, to decrease the effect of noise. Then, in combination with revised universal threshold, MNFNC uses chaotic particle swarm optimization algorithm to find the optimal values of parameters γ and α, which come from neighboring expansion and universal thresholds, respectively, in order to reconstruct the processed optimal wavelet coefficients to the original signal. We compare our proposed method with de-noising methods of other thresholding function, and the simulation results show MNFNC can improve the signal-to-noise ration, reduce distortion of the useful signal and effectively eliminate noise.