In this Letter, free-space optical (FSO) communication using patterned modulation and bucket detection is introduced to improve the bit error rate (BER) performance in complex and noisy environments. The scattered light is averaged in this communication structure. Second-order correlation, wavelet normalization, and compressed sensing are combined in the reconstruction algorithm. A signal with $N$ bits is reconstructed well from much less than $N$ measurements. Numerical simulations and experiments are performed without the narrowband optical filters used in traditional FSO communication. It can also be employed in real networks where secure communication is required. This provides the great opportunity to pave the way for real applications of FSO communication.