In MIMO-OFDM systems, the base station sends pilot signals for channel estimation at receivers. Since the number of pilots is proportional to the number of transmit antennas, the pilot overhead of massive MIMO-OFDM can be prohibitively large. Existing works have reduced pilot overhead by utilizing the sparsity and time correlation of channel impulse responses. In addition to these, we use the fact that the channel impulse response vector for each receive antenna has common sparsity due to the limited scattering around the base station in the massive MIMO systems. We present the joint channel estimation for all receive antennas based on compressive sensing techniques by utilizing the temporal correlation and spatial common sparsity. The proposed algorithm brings forth significant pilot reduction compared to the existing work while still achieving the same channel estimation performance.