Channel estimation is conceived for optical wireless scattering channels associated with Laser Diode transmitters and photon-counting/photomultiplier tube receivers. The proposed channel estimation approach consists of two stages, namely of the estimation of the channel tap second-order moments followed by the estimation of the channel taps based on the estimate of second-order moments. In the first stage, we provide the general framework of the moment estimation complemented by the conception of an estimation approach based on a sparse pilot structure, as well as by the analysis of the estimation error. We also propose the sparse pilot design and the associated low-complexity channel estimation, and prove the optimality of the proposed channel estimation. In the second stage, we conceive the channel tap estimation based on the eigenvalue decomposition of the matrix of estimated second-order moments, and analyze the associated performance. It is shown that as the length of the pilot sequence tends to infinity, the probability of having an estimation distortion above a threshold can be arbitrarily small. Simulation results show that the proposed sparse pilot sequence can lead to a smaller estimation error than the pilot design using random 0-1 bits.