An information feedback deviation-weighted method based on light intensity autocorrelation function (ACF) reconstruction is proposed to make full use of the effective particle size distribution (PSD) information in the decay section of ACF. The deviation-weighted inversion is carried out successively and the next deviation is reduced until the defined minimum information deviation is reached, that is, the distribution-reconstructed ACF obtained by inversion tends to be consistent with that obtained from the photon correlator. The inversion of the simulated data of the broad distribution and closely spaced bimodal distribution granular system at different noise levels is conducted. The results show that, compared with the routine weighting inversion methods, the proposed method can be used to obtain more accurate inversion results for the broad distribution and the closely spaced bimodal distribution and a better anti-noise performance is demonstrated, which are verified by the inversion results of the actual measurement data of standard polystyrene latex particles.