Quasi-cyclic low-density parity-check (QC-LDPC) codes have excelle nt performance,but in order to obtain the optimal performance on a specific channel,the decoding method needs to be modified to ad apt to the corresponding channel model.In this paper, we apply QC-LDPC codes into the channel based on avalanche photodiode (APD) rec eiver in the space optical communication system.The main work is to analyze the error performance and modi fy the belief propagation (BP) iterative decoding algorithm that uses probability density function (PDF) of the output sequences i n the multiplicative noise channel model.We introduce a new method to compute the log-likelihood ratio (LLR) init ialization for the LLR-BP decoding in the APD receiver channel.The simulation results show that the adaptation of QC-LDPC decoding on the accurate Webb-Gaussian model has better performance than that on approximate asymmetric Gaussian model,and almost no additional hardware complexity is increased.Moreover,the former can bring more than 1dB net coding gain compared with the latter at the bit error r ate (BER) of 10－6 under ideal channel conditions.