As the classical signal-to-noise rat io (SNR) estimation algorithms based on additive white Gaussian noise (AWGN) model can′t be applied to multiplicative noise chann el in optical communication systems,the effects of SNR mismatch on low-density parity-check (LDPC) decoding performance are analy zed in this paper.The simulation results prove that accurate SNR estimation is also necessary in the multiplicative noise channel for optical communication systems.An SNR estima tion algorithm based on quantization LDPC decision feedback is proposed.The rece ived optical signal is quantified,and the quantization SNR could be e stimated rough ly by the simplified expectation maximization (EM) algorithm and hard decision from LDPC decoding,and then the iterative LDPC decision feedback is us ed as accurate SNR estimation.The simulation results show that iterative decoding-aided process of LDPC codes can be accompl ished effectively without known parameters of the mean of the received signal and the noise variance.What′s more,after the SN R is estimated with the proposed method,there is only about 0.12dB loss for the LDPC decoding performance at the bit error rate (BER) of 10－5 compared with the ideal condition.