A photon-counting-based iterative parallel interference cancellation (PIC) scheme for free-space optical communications in the presence of multiple-access interference, shot noise, background radiation, and turbulence fading is designed. An efficient chip-level iterative equivalent noise estimation algorithm is also derived. Simulation results show that the proposed scheme can achieve a single-user performance, bound with the fast convergence property. More importantly, it can eliminate the bit-error rate floor of the conventional optical code-division multiple-access system with the aid of a relatively short spreading code length.OCIS Free-space optical (FSO) systems have received considerable attention due to their substantial bandwidth, low power consumption, and high security property [1,2] . The multiple-access interference (MAI) is the primary limitation that will strongly degrade system performance in the multi-user FSO communications [3] . The optical code-division multiple-access (OCDMA) has been developed as a promising candidate for highrate communications. OCDMA allows multiple users to share the same media simultaneously. The nonzero cross correlation for conventional non-coherent OCDMA systems can cause serious MAI [4] . Long optical orthogonal code (OOC) sequences are required to serve several users simultaneously, resulting in low bandwidth efficiency. Moreover, the complexity of multi-user detection (MUD) is always a serious concern.A photon-counting-based iterative parallel interference cancellation (Iter-PIC) for the FSO multi-user communications is designed based on the aforementioned scenarios. Each user can be separated by the user-specific interleaver, which avoids the complex design of OOC in OCDMA systems. A log-likelihood ratio (LLR) MAI Iter-PIC algorithm is derived at the receiver, which considers the shot noise, background radiation, and thermal noise. The bit-error rate (BER) performance is evaluated over Gamma-Gamma turbulence-induced fading channels.The Iter-PIC FSO system structure is shown in Fig.