In the large inversion range, wavelet-regularization inversion method (WRIM) is an effective method to improve the inversion accuracy of dynamic light scattering (DLS) data. However, the inversion accuracy of WRIM is still low for the strong noise and bimodal particles data. Based on WRIM, combined with the advantages of the traditional single scale Tikhonov and Truncated singular value decomposition (TSVD) regularization in DLS inversion, the multi-scale Tikhonov-TSVD-WRIM (TTWRIM) used for the DLS data is proposed. In this method, Tikhonov is applied to the adaptive adjustment of the coarse scale inversion range, and then TSVD is applied to the fine scale inversion, and its inversion result is processed by cubical smoothing algorithm with five-point approximation. Under three kinds of noise levels with 0.001, 0.005 and 0.01, the simulated data are inverted by Tikhonov, TSVD, WRIM and TTWRIM, respectively. The results show that TTWRIM has a high accuracy, strong anti-interference ability and bimodal resolution. Finally, the inversions of the measured particles also verify the conclusions of the simulated data. In the inversion range of ［1 nm, 2000 nm］, the peak value error of TTWRIM for 300 nm unimodal and 100~500 nm bimodal measured particles are 0.18% and 2.81%, respectively.