In this paper, two kinds of commonly used particle size inversion methods, regularization algorithm and Chahine algorithm, were used to inverse the simulated dynamic light scattering data of the unimodal distribution of 90nm and 250nm, the bimodal distribution of 50nm and 200 nm, and the bimodal distribution of 100 nm and 300 nm of the particles in submicron region, and measured the dynamic light scattering data of 105nm and 300nm particles, for comparing the noise effects of the two algorithms. The inversion results show that, the noise level is one of the key factor can restrict the accurate inversion for particle size measurement. The accuracy of inversion results decreases with the increase of the noise level, and when the noise level increase to a certain threshold value, the meaningful inversion results will not be obtained. Different inversion methods have different anti-noise ability, but there is no significant difference of the inversion results when noise level is very low. With the increase of the noise level, the retrieval results show a substantial difference: regularization method can effectively restrain the noise influence by the appropriate choice of the regularization parameter, which shows a better anti noise capability than Chahine algorithm. Compared with the Chahine algorithm, the regularization method, despite the need for regularization parameter setting, it does not need to assume the initial distribution. Therefore, it is recommended to use in noisy environment.